# deep-research

Deep research skill — broad parallel web searches, multi-source validation, confidence tracking, cited Markdown report. Supports 11 research types: market (TAM/SAM, segments, pricing, trends), domain (industry structure, ecosystem, regulatory landscape), technical (architecture, tools, benchmarks), competitive (competitor teardown, positioning, win/loss), product (feature analysis, reviews, roadmap signals), academic (literature survey, citation networks, key authors), person/org (due diligence on a company or public figure), financial (funding rounds, valuation multiples, revenue signals), legal (IP, patents, litigation, compliance), trend (emerging signals, foresight, scenario mapping), community (ecosystem health, key voices, governance, fragmentation). Use when asked to: 'research <topic>', 'deep dive on X', 'analyze the landscape', 'competitive analysis', 'compare these options', 'who are the players in Z', 'literature review', 'background on Y', 'what papers exist on X', 'product teardown', 'technology evaluation', 'regulatory overview', 'funding landscape', 'what trends are emerging in X', 'patent landscape', 'community health', or any request requiring scanning many sources and producing a cited written analysis. Apply whenever the deliverable is a thorough, sourced report rather than a quick answer. Trigger even when phrased casually: 'look into X', 'what's the deal with Y', 'dig into Z', 'I need to understand the space', 'catch me up on X'.

- **Kind:** skill
- **Source:** https://github.com/samber/cc-skills
- **Page:** https://forefy.com/skills/b3ff386b-09f0-4fa0-b63e-3585c3583561
- **API (JSON + files):** https://forefy.com/api/asr/b3ff386b-09f0-4fa0-b63e-3585c3583561

---

## SKILL.md

---
name: deep-research
description: "Deep research skill — broad parallel web searches, multi-source validation, confidence tracking, cited Markdown report. Supports 11 research types: market (TAM/SAM, segments, pricing, trends), domain (industry structure, ecosystem, regulatory landscape), technical (architecture, tools, benchmarks), competitive (competitor teardown, positioning, win/loss), product (feature analysis, reviews, roadmap signals), academic (literature survey, citation networks, key authors), person/org (due diligence on a company or public figure), financial (funding rounds, valuation multiples, revenue signals), legal (IP, patents, litigation, compliance), trend (emerging signals, foresight, scenario mapping), community (ecosystem health, key voices, governance, fragmentation). Use when asked to: 'research <topic>', 'deep dive on X', 'analyze the landscape', 'competitive analysis', 'compare these options', 'who are the players in Z', 'literature review', 'background on Y', 'what papers exist on X', 'product teardown', 'technology evaluation', 'regulatory overview', 'funding landscape', 'what trends are emerging in X', 'patent landscape', 'community health', or any request requiring scanning many sources and producing a cited written analysis. Apply whenever the deliverable is a thorough, sourced report rather than a quick answer. Trigger even when phrased casually: 'look into X', 'what's the deal with Y', 'dig into Z', 'I need to understand the space', 'catch me up on X'."
user-invocable: true
license: MIT
compatibility: Designed for Claude Code, Codex or similar harness. Requires internet access (web search and page fetching).
metadata:
  author: samber
  authors:
    - Maxme Courant (github.com/mcourant)
    - Samuel Berthe (github.com/samber)
  version: "1.2.0"
  openclaw:
    emoji: "🔎"
    homepage: https://github.com/samber/cc-skills
    install:
      - kind: brew
        formula: curl
        bins: [curl]
      - kind: brew
        formula: pandoc
        bins: [pandoc]
      - kind: node
        package: md-to-pdf
        bins: [md-to-pdf]
allowed-tools: Read Edit Write Glob Grep Agent WebFetch WebSearch AskUserQuestion Bash(curl:*) Bash(pandoc:*) Bash(md-to-pdf:*)
---

**Persona:** You are a senior research analyst. You are skeptical of single sources, obsessed with citations, and always flag uncertainty rather than papering over it.

**Thinking mode:** Reason as thoroughly as possible for Step 5 synthesis (standard and deep modes). Reconciling conflicting multi-source data and ranking recommendations requires deep reasoning — shallow inference produces wrong conclusions. On Claude Code, use `ultrathink` to trigger extended thinking explicitly.

**Modes:**

| Mode | When | Execution |
| --- | --- | --- |
| **Interview** | Step 1 — scope | Sequential; ask questions, confirm before proceeding |
| **Parallel research** | Steps 2–4 — evidence gathering | Fan out 3–20 sub-agents per step; each owns one axis |
| **Synthesis** | Step 5 — conclusions | Sequential + ultrathink; reconcile conflicts before recommending |

**Research depth** — select automatically based on the request:

| Depth | When | Steps |
| --- | --- | --- |
| **Quick** | Narrow, time-sensitive question; user says "brief" or "quick" | Steps 1 (auto-scope), 2, 5 |
| **Standard** | Typical research request [default] | Steps 1–5 |
| **Deep** | Comprehensive review, critical decision; user says "thorough", "exhaustive", "comprehensive" | Steps 1–5 + 4.5 (outline refinement) + critique pass |

**Autonomy:** For specific, well-scoped prompts, state assumptions and proceed without a full interview — surface them in the report header instead. Reserve the full scope interview for genuinely vague prompts (e.g., "Research blockchain", "Tell me about AI").

**Questions:** Ask the user through the environment's question tool — never as plain-text prose. One question at a time, 2–4 tappable options, wait for the answer. If the environment has no question tool, ask in prose with the same options, one at a time.

## Critical rules

- Web search is the core capability of this skill. If the environment has no web access, halt immediately and tell the user.
- **Every claim must cite a source URL.** Unsourced assertions are not findings — they are guesses.
- Critical claims (market size, growth rates, competitive positioning...) require **2+ independent sources** or get `confidence: Low`.
- Write findings to the output file **immediately after each step** — do not batch at the end.
- Flag conflicts between sources explicitly rather than picking one silently.
- **Prose-first:** Write in full sentences and paragraphs (aim for ≥80% prose). Use bullets only for true lists — never as the primary content delivery. "The market reached $4.2B in 2024 [Source]" is better than "\* Market: $4.2B".
- **Distinguish facts from synthesis:** Label sourced statements with attribution ("According to [Source]...") and analytical conclusions with hedges ("This suggests...", "The pattern across sources indicates..."). Never present inference as fact.
- **Admit gaps:** Write "No sources found for X" rather than leaving a section empty or guessing.

## Reference files

Load these files at the steps indicated only — not all upfront.

| File                            | Load at                             |
| ------------------------------- | ----------------------------------- |
| `references/citations.md`       | Step 2 (before first search)        |
| `references/parallel-search.md` | Step 2 (before spawning sub-agents) |
| `references/market.md`          | Step 2, if type == market           |
| `references/domain.md`          | Step 2, if type == domain           |
| `references/technical.md`       | Step 2, if type == technical        |
| `references/competitive.md`     | Step 2, if type == competitive      |
| `references/product.md`         | Step 2, if type == product          |
| `references/academic.md`        | Step 2, if type == academic         |
| `references/org.md`             | Step 2, if type == person/org       |
| `references/financial.md`       | Step 2, if type == financial        |
| `references/legal.md`           | Step 2, if type == legal            |
| `references/trend.md`           | Step 2, if type == trend            |
| `references/community.md`       | Step 2, if type == community        |

## Step 1 — Scope

First, get today's date: `date +%Y-%m-%d`. Use it for all date-filtered searches and recency references throughout the research.

**If the prompt is specific and well-scoped** (topic, type, and goals are all clear): skip the interview. Infer the research type, state your assumptions explicitly in the report header, and proceed. Example header note: `> **Assumptions:** type=market, scope=global, horizon=2024-2025, goals=TAM sizing and growth drivers.`

**If the prompt is vague or ambiguous** (e.g., "Research blockchain", "Tell me about AI"): ask the user:

1. What type? (see list below)
2. What specific questions or goals should the research answer?
3. Any geographic, time, or segment constraints?

Research types:

- `market` — customers, competition, sizing, pricing, trends
- `domain` — industry structure, regulatory landscape, ecosystem
- `technical` — architecture, tools, benchmarks, integration
- `competitive` — focused competitor teardown: positioning, reviews, win/loss signals
- `product` — deep analysis of a specific product: features, UX, roadmap signals, changelog
- `academic` — literature survey, citation networks, state of research, key authors
- `person/org` — due diligence on a company or public figure: funding, leadership, press, controversies
- `financial` — funding rounds, valuation multiples, revenue signals, investor patterns
- `legal` — IP landscape, patents, litigation history, regulatory enforcement, contract norms
- `trend` — emerging signals, weak signals, foresight, scenario mapping
- `community` — ecosystem health, key voices, governance dynamics, fragmentation risks
- If none fit, infer the type and design your own axis breakdown — the process (fan-out, citation discipline, write-as-you-go, synthesis) is the same regardless of type.

Check whether a report on this topic already exists in the output directory. If found, summarize what it covers and ask: extend or start fresh?

Set output path: `./research/{type}-{topic}-{YYYY-MM-DD}.md` (lowercase, hyphens). Ask if the user wants a different path. Load `assets/report-template.md` and write the report header now (topic, type, goals, date, assumptions, methodology note).

## Step 2 — Core research (parallel fan-out)

Load `references/citations.md` and `references/parallel-search.md`. Load the type-specific reference file.

Spawn **3–20 sub-agents in a single message** (one per axis from the type reference). Each agent:

- Searches its axis on the web and fetches the sources it cites
- Writes findings as prose paragraphs with inline citations — not bullet lists
- Returns URL, accessed date, and confidence level per claim
- Tags each source: **Primary** (official docs, filings, peer-reviewed), **Established** (major publications, analyst firms), or **Low** (blogs, forums, single opinions). Flag Low-tier sources prominently.
- Does not wait for other agents

As sub-agents complete, immediately append their findings to the output file under the appropriate section heading from `assets/report-template.md`. Do not wait for all agents to finish before writing.

## Step 3 — Competitive / landscape analysis (parallel fan-out)

Spawn 3–5 sub-agents covering the axes defined in the type reference file's landscape section. Same citation discipline. Append results to the output file immediately.

## Step 4 — Deep dive (parallel fan-out)

Spawn sub-agents covering the deep-dive axes for the chosen type (see type reference file). Append results immediately.

## Step 4.5 — Outline refinement (deep mode only)

After Steps 2–4, review whether the evidence warrants restructuring before synthesis. Ask:

- Did findings contradict the initial scope assumptions?
- Did an important angle emerge that wasn't in the original plan?
- Are any sections underpowered by evidence — or overloaded?

If yes: adapt the outline. Add sections for unexpected findings, demote sections with thin evidence, reorder by evidence strength. Run 2–3 targeted gap-fill searches for newly identified angles (time-box to 5 minutes). Document what changed and why in the report's methodology note.

Skip in quick and standard modes.

## Step 5 — Synthesis

**Use `ultrathink` here** (standard and deep modes).

Read the full output file. Write the synthesis section:

```md
## Key Findings

(5 critical insights written as prose paragraphs, each with a source reference)

## Strategic Recommendations

1. [Recommendation] — Rationale. Evidence: [source].
2. ... (3–5 recommendations, ranked by impact)

## Risks and Uncertainties

- Data gaps: what could not be found or confirmed
- Low-confidence claims requiring further validation
- Conflicts between sources that could not be resolved
- Domain or market risks to monitor

## Next Steps

- Recommended follow-up research
- If the initial request is not fulfilled, loop on step 1 and ask more questions
- Decisions this research enables
```

Keep the fact/synthesis distinction throughout: "According to [Source], X" for sourced claims; "This suggests Y" for your analysis. If a recommendation rests on Low-confidence data, say so explicitly.

**Critique pass (deep mode only):** Before finalizing, red-team the synthesis. Ask: What's missing? What could be wrong? What alternative explanations exist? What biases might be present? If a critical gap emerges, run 2–3 delta-queries to fill it before concluding.

## Step 6 — PDF export (optional)

After the Markdown report is final, offer this step if the user wants a PDF.

Try each tool in order, stop at the first that works:

1. **Pandoc** (best output quality):

   ```bash
   pandoc report.md -o report.pdf --pdf-engine=wkhtmltopdf
   # or with weasyprint:
   pandoc report.md -o report.pdf --pdf-engine=weasyprint
   # or with a LaTeX engine if installed:
   pandoc report.md -o report.pdf
   ```

2. **`md-to-pdf`** (Node, no LaTeX required):

   ```bash
   md-to-pdf report.md
   ```

Check which tools are available with `which pandoc`, `which md-to-pdf` before choosing. If neither is available, tell the user which to install.

## Pitfalls

- Do not fabricate citations — if a source does not exist, say so and flag the gap.
- Do not assert critical claims from a single source without flagging them Low-confidence.
- Do not batch findings — write to the file after each step, not at the end.
- Do not over-claim on Low-confidence data — hedge explicitly.
- Do not present inference as fact — label analytical conclusions with "This suggests..." or similar hedges.
- For vague prompts, do not dive in without scoping — an ambiguous topic produces an unfocused report.

## Disclaimer

Research reflects a snapshot in time. Web content changes. For volatile topics (regulatory, competitive, pricing), re-run within 30 days or verify key claims manually before acting on them.

## assets

```

```

## assets/report-template.md

---
topic: "{TOPIC}"
type: "{market|domain|technical|competitive|product|academic|person-org|financial|legal|trend|community}"
goals: "{GOALS}"
date: "{YYYY-MM-DD}"
methodology: "Parallel web research via sub-agents. Citations inline per references/citations.md. Confidence levels: High / Medium / Low."
---

# Research Report — {TOPIC}

> **Type:** {type} | **Date:** {YYYY-MM-DD} | **Constraints:** {CONSTRAINTS or "none"}
>
> **Goals:** {BULLET_GOALS}

---

<!-- Steps 2–4: findings are appended here as sub-agents complete. Do not leave this comment in the final report. -->

<!-- TYPE: market -->

## Market Overview

### Market Size and Growth

_Pending research._

### Customer Segments

_Pending research._

### Pricing Models

_Pending research._

### Trends and Dynamics

_Pending research._

## Competitive Landscape

### Key Players

_Pending research._

### Positioning Map

_Pending research._

### Competitive Matrix

| Player | Segment | Differentiator | Pricing | Strengths | Weaknesses |
| ------ | ------- | -------------- | ------- | --------- | ---------- |
| _TBD_  |         |                |         |           |            |

### Market Gaps

_Pending research._

## Customer Insights

### Top Pain Points

_Pending research._

### Purchase Triggers

_Pending research._

### Customer Journey

_Pending research._

<!-- END TYPE: market -->

<!-- TYPE: domain -->

## Domain Overview

### Industry Structure

_Pending research._

### Core Terminology

_Pending research._

### Current State and Trajectory

_Pending research._

### Established Best Practices

_Pending research._

## Ecosystem Map

### Supply and Demand Chain

_Pending research._

### Substitutes and Adjacent Domains

_Pending research._

### Tools and Platforms

_Pending research._

### Leading Organizations

_Pending research._

## Regulatory Landscape

### Current Regulations

_Pending research._

### Upcoming Changes

_Pending research._

### Compliance Requirements

_Pending research._

<!-- END TYPE: domain -->

<!-- TYPE: technical -->

## Technology Landscape

### Overview of Approaches

_Pending research._

### Performance and Benchmarks

_Pending research._

### Community Health and Maturity

_Pending research._

### Integration and Compatibility

_Pending research._

## Comparative Analysis

### Top Options Head-to-Head

_Pending research._

### Decision Matrix

| Option | Performance | DX  | Community | Cost | Lock-in | Notes |
| ------ | ----------- | --- | --------- | ---- | ------- | ----- |
| _TBD_  |             |     |           |      |         |       |

### Migration and Lock-in Risks

_Pending research._

## Implementation Considerations

### Recommended Architecture Patterns

_Pending research._

### Common Pitfalls and Gotchas

_Pending research._

### Security and Compliance

_Pending research._

<!-- END TYPE: technical -->

<!-- TYPE: competitive -->

## Competitor Profiles

### {Competitor A}

_Pending research._

### {Competitor B}

_Pending research._

## Comparison Matrix

| Competitor | Segment | Differentiator | Pricing | Strengths | Weaknesses |
| ---------- | ------- | -------------- | ------- | --------- | ---------- |
| _TBD_      |         |                |         |           |            |

## Cross-Competitor Analysis

### Review Sentiment

_Pending research._

### Pricing Comparison

| Competitor | Free tier | Entry | Mid | Enterprise |
| ---------- | --------- | ----- | --- | ---------- |
| _TBD_      |           |       |     |            |

### GTM and Distribution

_Pending research._

## Win/Loss Signals

### Switching Signals

_Pending research._

### Hiring Signals

_Pending research._

### Market Gaps

_Pending research._

<!-- END TYPE: competitive -->

<!-- TYPE: product -->

## Product Overview

### Target Audience

_Pending research._

### Feature Surface

_Pending research._

### Pricing and Packaging

_Pending research._

### Technology and Architecture

_Pending research._

### Integrations and Ecosystem

_Pending research._

## User and Market Perception

### Review Summary

_Pending research._

### Changelog Highlights (last 12 months)

_Pending research._

### Community Health

_Pending research._

## Strategic Signals

### Competitive Positioning

_Pending research._

### Adoption Signals

_Pending research._

### Risk Signals

_Pending research._

<!-- END TYPE: product -->

<!-- TYPE: academic -->

## Literature Overview

### Foundational Works

_Pending research._

### Recent Advances (last 2–3 years)

_Pending research._

### Key Authors and Institutions

_Pending research._

### Survey Papers and Reviews

_Pending research._

## Citation Landscape

### Citation Clusters and Sub-communities

_Pending research._

### Contested or Retracted Claims

_Pending research._

### Research-to-Practice Gap

_Pending research._

## State of the Field

### Open Problems

_Pending research._

### Datasets and Benchmarks

_Pending research._

### Top Venues

_Pending research._

<!-- END TYPE: academic -->

<!-- TYPE: person-org -->

## Profile

### Background

_Pending research._

### Financial Profile

_Pending research._

### Leadership

_Pending research._

### Products and Customers

_Pending research._

## Press and Perception

### Press Coverage

_Pending research._

### Controversies and Risks

_Pending research._

### Social Sentiment

_Pending research._

## Strategic Context

### Partnerships and Ecosystem

_Pending research._

### Hiring Signals

_Pending research._

### Competitive Position

_Pending research._

<!-- END TYPE: person-org -->

<!-- TYPE: financial -->

## Funding and Capital Activity

### Recent Funding Rounds

_Pending research._

### Investor Landscape

_Pending research._

## Revenue and Business Performance

### Revenue Signals

_Pending research._

### Unit Economics Signals

_Pending research._

### Capital Efficiency

_Pending research._

## Valuation Context

### Comparable Multiples

| Company / Sector | Metric | Multiple | Source | Year |
| ---------------- | ------ | -------- | ------ | ---- |
| _TBD_            |        |          |        |      |

### Public Market Comparables

_Pending research._

## M&A and Exit Landscape

_Pending research._

## Macro and Risk Context

### Macro Environment

_Pending research._

### Regulatory and Accounting Risk

_Pending research._

<!-- END TYPE: financial -->

<!-- TYPE: legal -->

## Intellectual Property Landscape

### Key Patents

_Pending research._

### Trademarks and Brand Protection

_Pending research._

### Copyright and Licensing Norms

_Pending research._

### IP Disputes

_Pending research._

## Regulatory Enforcement and Compliance

### Enforcement Actions (last 5 years)

_Pending research._

### Compliance Requirements by Jurisdiction

_Pending research._

### Upcoming Regulatory Changes

_Pending research._

## Litigation and Contract Norms

### Litigation History

_Pending research._

### Standard Contract Terms

_Pending research._

### Jurisdictional Risk

_Pending research._

> ⚠️ This report surfaces publicly available legal information for research purposes only. It does not constitute legal advice.

<!-- END TYPE: legal -->

<!-- TYPE: trend -->

## Signal Detection

### Analyst and Research Forecasts

_Pending research._

### Investment Signals

_Pending research._

### Academic and R&D Signals

_Pending research._

### Cultural and Behavioral Signals

_Pending research._

## Momentum and Adjacent Forces

### Technology Enablers

_Pending research._

### Adjacent Domain Spillovers

_Pending research._

### Regulatory and Geopolitical Forces

_Pending research._

## Scenario Map

### Bear Case

_Pending research._

### Base Case (consensus trajectory)

_Pending research._

### Bull Case

_Pending research._

<!-- END TYPE: trend -->

<!-- TYPE: community -->

## Community Profile

### Size and Reach

_Pending research._

### Key Voices and Leaders

_Pending research._

### Core Platforms and Venues

_Pending research._

### Content and Knowledge Production

_Pending research._

## Health and Dynamics

### Activity and Engagement Trends

_Pending research._

### Governance and Power Dynamics

_Pending research._

### Tone and Culture

_Pending research._

## Strategic Signals

### Commercial Activity

_Pending research._

### Fragmentation and Competing Communities

_Pending research._

### External Dependencies and Risks

_Pending research._

<!-- END TYPE: community -->

---

## Key Findings

_Populated at Step 5._

## Strategic Recommendations

_Populated at Step 5._

## Risks and Uncertainties

_Populated at Step 5._

## Next Steps

_Populated at Step 5._

## evals

```

```

## evals/evals.json

```json
{
  "skill_name": "deep-research",
  "metadata": {
    "eval_methodology": "adversarial — each eval has a trap the model falls into without the skill",
    "without_skill_runner_note": "When running without-skill evals, disable ALL skills from this plugin. The deep-research skill enforces citation discipline, confidence tracking, and source conflict surfacing that the model does not apply by default."
  },
  "evals": [
    {
      "id": 1,
      "name": "citation-enforcement",
      "prompt": "Research the global cybersecurity market: current size, growth projections for the next 5 years, and the top 3 demand drivers. I need a sourced report.",
      "trap": "Without the skill, the model produces a report with numeric claims (market size, CAGR) but omits inline citation URLs, accessed dates, and confidence levels. It may also pick one analyst estimate and present it as fact rather than surfacing conflicting figures.",
      "expected_output": "A markdown report where every numeric claim (market size, CAGR, projections) has an inline citation with a URL, an accessed date, and a confidence level. Conflicting analyst estimates are surfaced and compared rather than silently picking one. No figure is stated without a source.",
      "files": [],
      "assertions": [
        { "id": "1.1", "description": "every numeric claim includes an inline citation with a source URL" },
        { "id": "1.2", "description": "every citation includes an accessed date in YYYY-MM-DD format" },
        { "id": "1.3", "description": "every citation includes a confidence level (High, Medium, or Low)" },
        { "id": "1.4", "description": "conflicting analyst estimates are surfaced and compared rather than silently picking one" },
        { "id": "1.5", "description": "no market size or CAGR figure is stated without a named source" },
        { "id": "1.6", "description": "findings are written as prose paragraphs, not bullet lists" }
      ]
    },
    {
      "id": 2,
      "name": "confidence-private-company",
      "prompt": "What is Stripe's current annual revenue and valuation? I want a financial research report.",
      "trap": "Without the skill, the model states Stripe revenue and valuation figures from training data as if they were verified facts, without flagging that Stripe is private and has no public filings. It typically picks one reported figure and presents it authoritatively.",
      "expected_output": "A markdown report where all revenue and valuation figures are explicitly marked as reported/unverified (confidence: Medium or Low), since Stripe is a private company with no public filings. The report cites press sources, not official filings, and clearly hedges on figure accuracy. Multiple reported estimates are compared where available.",
      "files": [],
      "assertions": [
        { "id": "2.1", "description": "explicitly states that Stripe is a private company with no public financial filings" },
        { "id": "2.2", "description": "all revenue and valuation figures are labeled as reported or unverified (not presented as fact)" },
        { "id": "2.3", "description": "confidence level is Medium or Low for all financial figures" },
        { "id": "2.4", "description": "cites press sources rather than claiming official filings as the source" },
        { "id": "2.5", "description": "compares multiple reported estimates rather than picking one authoritative number" }
      ]
    },
    {
      "id": 3,
      "name": "source-conflict-surfacing",
      "prompt": "What is the current global market size of the plant-based food market? Research this and give me the figures.",
      "trap": "Without the skill, the model picks a single analyst figure and states it as the market size, without noting the wide discrepancy between estimates that results from differing scope definitions.",
      "expected_output": "A markdown report that surfaces the wide discrepancy between analyst estimates (figures range from ~$8B to $160B+ depending on definition and scope). Conflicts between sources are explicitly flagged, methodology differences identified as the root cause, and a conservative directional estimate stated rather than picking one authoritative number.",
      "files": [],
      "assertions": [
        { "id": "3.1", "description": "surfaces the wide discrepancy between analyst estimates (multiple different figures cited)" },
        { "id": "3.2", "description": "explicitly flags that conflicting estimates exist rather than silently picking one" },
        { "id": "3.3", "description": "identifies methodology or scope differences as the root cause of the discrepancy" },
        { "id": "3.4", "description": "does not present a single figure as the definitive market size" },
        { "id": "3.5", "description": "states a directional or conservative estimate with appropriate hedging" }
      ]
    },
    {
      "id": 4,
      "name": "scope-interview-vague",
      "prompt": "Research blockchain.",
      "trap": "Without the skill, the model dives into a generic blockchain overview (history, consensus mechanisms, use cases) without scoping the request. With the skill, it recognizes this as vague, either asks clarifying questions or explicitly states assumptions before proceeding.",
      "expected_output": "The report documents that the prompt is too vague to research without scoping — it lists the clarifying questions that would normally be asked (type, goals, constraints, geographic focus, time horizon), states explicit assumptions before proceeding, and produces a scoped report rather than a generic blockchain overview.",
      "files": [],
      "assertions": [
        { "id": "4.1", "description": "recognizes the prompt as vague or ambiguous rather than diving in immediately" },
        { "id": "4.2", "description": "either asks clarifying questions or explicitly states scoping assumptions before producing the report" },
        { "id": "4.3", "description": "the stated assumptions cover at least: research type, goals, and constraints" },
        { "id": "4.4", "description": "produces a scoped report rather than a generic blockchain overview" },
        { "id": "4.5", "description": "does not present inference or training-data summaries as sourced research findings" }
      ]
    },
    {
      "id": 5,
      "name": "legal-type-axis-coverage",
      "prompt": "What does a B2B SaaS company need to know about GDPR compliance when handling EU user data? Research the current requirements, recent enforcement actions, and upcoming regulatory changes.",
      "trap": "Without the skill, the model gives a general GDPR summary from training data without structuring around legal research axes, without citing official EU sources, and without including enforcement actions with specific fine amounts and companies.",
      "expected_output": "A structured markdown report organized around the legal research axes: current regulations (with official EU sources), recent enforcement actions with fine amounts and offending companies cited, upcoming changes (e.g., DMA, AI Act interactions), and specific compliance requirements enumerated. Includes a legal disclaimer. Sources from EUR-Lex, EDPB, or reputable legal publishers.",
      "files": [],
      "assertions": [
        { "id": "5.1", "description": "report is structured around legal axes: IP/regulations, enforcement, compliance requirements, upcoming changes" },
        { "id": "5.2", "description": "enforcement actions include specific fine amounts and named companies" },
        { "id": "5.3", "description": "upcoming regulatory changes are mentioned (e.g., AI Act, DMA interactions with GDPR)" },
        { "id": "5.4", "description": "includes a legal disclaimer stating this is not legal advice" },
        { "id": "5.5", "description": "cites official sources such as EUR-Lex, EDPB, or DPA decisions rather than only blog posts" }
      ]
    },
    {
      "id": 6,
      "name": "pdf-download-attempt",
      "prompt": "Research the key findings of the IPCC Sixth Assessment Report (AR6) on climate risks, adaptation strategies, and the 1.5°C vs 2°C warming scenarios. I want specific data points, not a general summary.",
      "trap": "Without the skill, the model paraphrases training data about IPCC AR6 as if it were a sourced finding, without attempting to fetch the actual documents and without chapter-level attribution for specific figures.",
      "expected_output": "The report attempts to download the IPCC AR6 PDF via curl, cites specific chapters and sections (not just 'the IPCC says'), includes precise figures with chapter-level attribution, and goes beyond training data paraphrasing with direct URL references to the official ipcc.ch documents.",
      "files": [],
      "assertions": [
        { "id": "6.1", "description": "attempts to download or fetch the IPCC AR6 document via curl or WebFetch rather than relying on training data" },
        { "id": "6.2", "description": "cites specific chapters and sections of AR6 rather than attributing everything to 'the IPCC'" },
        { "id": "6.3", "description": "includes precise numeric figures with chapter-level attribution" },
        { "id": "6.4", "description": "references official ipcc.ch URLs rather than only secondary sources" },
        { "id": "6.5", "description": "distinguishes sourced claims from synthesis using 'According to [chapter]...' attribution" }
      ]
    }
  ]
}
```

## references

```

```

## references/academic.md

# Academic / Literature Research — Axes, Probes, and Section Template

Use this file for `type == academic` research: surveying academic literature, mapping citation networks, identifying key authors and institutions. Load after the user confirms scope.

Probe the user: "What field or sub-field? Any time range (e.g., last 5 years)? Are you looking for foundational papers, recent advances, or the current consensus on a specific claim?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Literature mapping

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Foundational works** | What are the seminal papers? Most-cited works? When were they published and by whom? Search Google Scholar, Semantic Scholar, arXiv |
| 2 | **Recent advances** | Papers from the last 2–3 years on the topic. What new methods, findings, or frameworks have emerged? |
| 3 | **Key authors and labs** | Who publishes most in this area? Which institutions? Any dominant research groups or labs? |
| 4 | **Surveys and reviews** | Are there published survey papers or meta-analyses? These are high-leverage sources that summarize a field |

## Step 3 axes — Citation landscape

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Citation clusters** | Which papers are frequently cited together? What sub-communities exist within the broader topic? |
| 2 | **Contested claims** | Where do papers disagree? What findings have been challenged, retracted, or failed to replicate? |
| 3 | **Practitioner adoption** | Which academic findings have been adopted in industry or open-source tools? Where is there a gap between research and practice? |

## Step 4 axes — State of the field

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Open problems** | What do recent papers identify as unsolved? What are the stated next research directions? |
| 2 | **Datasets and benchmarks** | What standard datasets or evaluation benchmarks does the field use? Who maintains them? |
| 3 | **Conferences and venues** | Which conferences and journals are considered top-tier? Acceptance rates? Recent best papers? |

## Citation notes for academic research

Academic citations follow the same URL + confidence format, but also include:

- DOI when available: `[Author et al. (Year)](https://doi.org/...)`
- arXiv IDs for preprints: `[Author et al. (Year)](https://arxiv.org/abs/XXXX.XXXXX)`
- Distinguish peer-reviewed publications (High) from preprints (Medium) from blog posts summarizing papers (Low)

## Section template

```md
## Literature Overview

### Foundational Works

### Recent Advances (last 2–3 years)

### Key Authors and Institutions

### Survey Papers and Reviews

## Citation Landscape

### Citation Clusters and Sub-communities

### Contested or Retracted Claims

### Research-to-Practice Gap

## State of the Field

### Open Problems

### Datasets and Benchmarks

### Top Venues
```

## references/citations.md

# Citation Rules

Every claim in the report must be traceable to a source. This file defines the format and confidence ladder.

## Inline citation format

Append a citation immediately after the claim:

```
[Source Name](https://url) (accessed YYYY-MM-DD, confidence: High|Medium|Low)
```

**Short source name rules:**

- Use the publication or domain name, not the article title: `Gartner`, `TechCrunch`, `GitHub`, `NIST`, `SEC.gov`
- For blog posts or forums: use author full name or handle + domain: `@jsmith / HN`, `Stack Overflow`
- For primary sources (official docs, filings): use the org name: `Apple SEC 10-K`, `EU Commission`

## Confidence levels

| Level | Meaning |
| --- | --- |
| **High** | 2+ reputable, independent sources agree on the claim |
| **Medium** | 1 reputable source (established publication, official document, primary source) |
| **Low** | Blog post, forum, single analyst opinion, or inferred from indirect data |

A source is _reputable_ if it has an editorial or review process, institutional backing, or is a primary source (government filings, official docs, peer-reviewed research).

## Multi-source validation for critical claims

Critical claims — market size, growth rates, competitive market share, regulatory deadlines — **must** reach High confidence or be explicitly flagged:

```
> ⚠️ **Confidence: Low** — Only one source found for this figure. Treat as directional until corroborated.
```

If two sources conflict, do not pick one silently. Report the conflict:

```
Sources disagree: [Source A](url-a) reports $4.2B; [Source B](url-b) reports $6.1B.
Likely difference: methodology (TAM vs SAM). Using the lower figure as conservative baseline.
```

## Accessed date

Always include the access date. Research pages change. A figure cited today may not match tomorrow.

Format: `YYYY-MM-DD` (ISO 8601).

## What counts as a source

- ✓ News articles from established outlets
- ✓ Analyst reports (Gartner, IDC, Forrester, CB Insights — even paywalled summaries)
- ✓ Government or regulatory publications
- ✓ Official product documentation or release notes
- ✓ Peer-reviewed papers
- ✓ Company investor filings (10-K, S-1, earnings calls)
- ✗ Wikipedia (use it to find primary sources, not as a source itself)
- ✗ Aggregator lists without original research
- ✗ Social media posts (unless the author is the primary source, e.g., a founder's tweet about their own product)

## references/community.md

# Community / Ecosystem Research — Axes, Probes, and Section Template

Use this file for `type == community` research: health and dynamics of a community or ecosystem — professional, consumer, open source, geographic, industry association, fan, or brand community. Load after the user confirms scope.

Probe the user: "What kind of community — professional network, consumer brand, open-source project, industry association, geographic cluster, or something else? What do you want to understand — size, health, key voices, dynamics, or risks?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

This type is intentionally broad. Adapt the axes to the community kind. Not all axes apply to every community.

## Step 2 axes — Community profile and size

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Size and reach** | Membership counts, follower numbers, active participant estimates across platforms (LinkedIn groups, Reddit, Discord, Slack, forums, mailing lists, local meetup attendance) |
| 2 | **Key voices and leaders** | Who are the recognized authorities, moderators, or informal leaders? How did they earn credibility? |
| 3 | **Core platforms and venues** | Where does the community actually gather? Which platforms are primary vs secondary? Conference or events calendar |
| 4 | **Content and knowledge production** | Newsletters, podcasts, blogs, YouTube channels, journals, or reports the community produces and relies on |

## Step 3 axes — Health and dynamics

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Activity and engagement** | Is the community growing or declining? Post frequency, event attendance trends, content publishing cadence |
| 2 | **Governance and power dynamics** | Who controls the community? Corporate-backed vs independent? Formal governance (bylaws, board) vs informal? Any power conflicts or factions? |
| 3 | **Tone and culture** | What are the stated and unstated norms? Welcoming to newcomers? Tolerance for dissent? Recurring tensions or taboos |

## Step 4 axes — Strategic signals

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Commercial activity** | Sponsorships, job boards, vendor activity, paid events, monetization models — and community's attitude toward commercial participation |
| 2 | **Fragmentation and splinter communities** | Are there competing communities covering the same topic? Any notable schisms or forks? |
| 3 | **External dependencies and risks** | Platform dependency risks (e.g., a community hosted entirely on a single corporate Slack). Any existential threats — platform shutdowns, key person departure, topic obsolescence |

## Notes on open-source communities

For open-source projects specifically, supplement with:

- Contributor concentration (bus factor): how many people could leave and break the project?
- Commit activity over time (GitHub Insights or similar)
- Issue and PR response time as a governance health signal
- Licensing clarity and any license change history

## Section template

```md
## Community Profile

### Size and Reach

### Key Voices and Leaders

### Core Platforms and Venues

### Content and Knowledge Production

## Health and Dynamics

### Activity and Engagement Trends

### Governance and Power Dynamics

### Tone and Culture

## Strategic Signals

### Commercial Activity

### Fragmentation and Competing Communities

### External Dependencies and Risks
```

## references/competitive.md

# Competitive Research — Axes, Probes, and Section Template

Use this file for `type == competitive` research: focused teardown of competitors, without the full market-sizing context of a market research report. Load after the user confirms scope.

Probe the user: "Which competitors? Any specific dimensions to compare (pricing, features, GTM, hiring, tech stack)?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Competitor profiles

Fan out one sub-agent **per competitor** (up to 5), each producing a structured profile:

| Agent | Axis | Key questions |
| --- | --- | --- |
| Per competitor | **Company profile** | Founding year, headcount, funding raised, investors, revenue signals (if public), business model |
| Per competitor | **Positioning & messaging** | Website headline, target persona, claimed differentiator, tone |
| Per competitor | **Product surface** | Feature set, pricing tiers, integrations, what's on the roadmap (if public) |
| Per competitor | **Community signals** | GitHub stars/activity (if applicable), forum presence, job postings, developer sentiment |

## Step 3 axes — Cross-competitor analysis

Fan out these 3 axes (one agent each):

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Reviews & sentiment** | G2, Capterra, Reddit, HN, App Store: recurring praise and recurring complaints per competitor |
| 2 | **Pricing comparison** | Side-by-side pricing tiers; free tier vs paid; pricing page changes over time (Wayback Machine) |
| 3 | **GTM & distribution** | How do they acquire customers? SEO, paid, PLG, sales-led, partnerships? Content strategy? |

## Step 4 axes — Win/loss signals

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Switching signals** | Posts/threads about people switching away from each competitor; stated reasons |
| 2 | **Hiring signals** | Recent job postings; what they reveal about strategic bets (e.g., "enterprise AE" = moving upmarket) |
| 3 | **Market gaps** | What no competitor does well? Underserved segments, missing features, pricing gaps |

## Section template

```md
## Competitor Profiles

### {Competitor A}

### {Competitor B}

...

## Cross-Competitor Analysis

### Review Sentiment

### Pricing Comparison

| Competitor | Free tier | Entry | Mid | Enterprise |
| ---------- | --------- | ----- | --- | ---------- |

### GTM and Distribution

## Win/Loss Signals

### Switching Signals

### Hiring Signals

### Market Gaps
```

## references/domain.md

# Domain Research — Axes, Probes, and Section Template

Use this file for `type == domain` research (industry analysis, regulatory landscape, ecosystem mapping). Load after the user confirms scope.

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Core domain research

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Industry structure** | Who are the key players (producers, distributors, buyers, intermediaries)? Concentration vs fragmentation? |
| 2 | **Terminology & concepts** | What jargon and acronyms does a practitioner use? What are the foundational concepts a newcomer must know? |
| 3 | **Current state & trajectory** | Where is the domain today vs 5 years ago? What major shifts happened? What's the consensus view of where it's heading? |
| 4 | **Best practices & standards** | What approaches are considered established best practice? Any formal standards bodies or certifications? |

## Step 3 axes — Ecosystem mapping

Fan out these 4 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Suppliers & buyers** | Who supplies the domain's inputs? Who consumes its outputs? Power dynamics? |
| 2 | **Substitutes & alternatives** | What could replace this domain or make it irrelevant? Adjacent domains competing for the same problem? |
| 3 | **Tools & platforms** | What software, platforms, or infrastructure do practitioners rely on? Any dominant standards? |
| 4 | **Leading organizations** | Who sets the agenda — industry associations, thought leaders, dominant firms, open-source foundations? |

## Step 4 axes — Regulatory & compliance deep dive

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Current regulations** | What laws, directives, or standards apply today? Jurisdiction? Enforcement track record? |
| 2 | **Upcoming changes** | What regulations are proposed, in consultation, or recently passed but not yet in force? |
| 3 | **Compliance requirements** | What must an operator actually do to comply? Certifications, audits, reporting obligations, data handling rules? |

## Section template

Use these headings in the output file:

```md
## Domain Overview

### Industry Structure

### Core Terminology

### Current State and Trajectory

### Established Best Practices

## Ecosystem Map

### Supply and Demand Chain

### Substitutes and Adjacent Domains

### Tools and Platforms

### Leading Organizations

## Regulatory Landscape

### Current Regulations

### Upcoming Changes

### Compliance Requirements
```

## references/financial.md

# Financial Research — Axes, Probes, and Section Template

Use this file for `type == financial` research: funding landscapes, valuation analysis, revenue signals, investor patterns, capital markets. Load after the user confirms scope.

Probe the user: "What's the focus — a specific company's financials, a sector's funding landscape, valuation multiples, or investor activity? Any time range?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

⚠️ Distinguish clearly between **public** (filed, verified) and **reported** (press-cited, unconfirmed) figures. Label each.

## Step 2 axes — Capital and revenue landscape

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Funding activity** | Recent funding rounds (Crunchbase, PitchBook, press): amounts, investors, dates, stated valuations. For a sector: total capital deployed, deal counts, stage distribution |
| 2 | **Revenue signals** | Public filings (10-K, S-1, earnings transcripts), press releases, analyst estimates, CEO statements. For private companies: reported ARR, headcount-to-revenue proxies |
| 3 | **Investor landscape** | Who are the active investors? Lead vs follow-on patterns, stage focus, portfolio overlaps, notable exits |
| 4 | **Valuation benchmarks** | Revenue multiples, EBITDA multiples, or GMV multiples for comparable companies or the sector. Source and year of each multiple |

## Step 3 axes — Financial health and trends

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Unit economics signals** | CAC, LTV, gross margin signals from public statements, job postings, or analyst commentary |
| 2 | **Capital efficiency** | Burn multiples, time to profitability signals, headcount vs revenue proxies, layoff history |
| 3 | **M&A and exit landscape** | Recent acquisitions in the space: acquirers, deal sizes, strategic rationale. IPO pipeline if relevant |

## Step 4 axes — Macro and risk context

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Macro environment** | Interest rate impact on valuations, sector-specific capital cycle (boom/correction?), LP appetite for the asset class |
| 2 | **Regulatory and accounting risk** | Pending regulations affecting valuation or revenue recognition, tax changes, foreign investment restrictions |
| 3 | **Public market comparables** | For public companies: P/E, EV/Revenue, EV/EBITDA vs sector peers. Recent stock performance and analyst ratings |

## Citation notes for financial research

Financial figures require extra rigor:

- Always cite the source type: public filing, earnings call, press release, analyst estimate, or reported (unverified)
- Reported figures from press: `confidence: Medium` at best — flag as unverified
- Multi-year figures: note the fiscal year, not just the date
- Currency: always specify (USD, EUR, etc.)

## Section template

```md
## Funding and Capital Activity

### Recent Funding Rounds

### Investor Landscape

## Revenue and Business Performance

### Revenue Signals

### Unit Economics Signals

### Capital Efficiency

## Valuation Context

### Comparable Multiples

| Company / Sector | Metric | Multiple | Source | Year |
| ---------------- | ------ | -------- | ------ | ---- |

### Public Market Comparables

## M&A and Exit Landscape

## Macro and Risk Context

### Macro Environment

### Regulatory and Accounting Risk
```

## references/legal.md

# Legal Research — Axes, Probes, and Section Template

Use this file for `type == legal` research: IP landscape, patent analysis, litigation history, regulatory enforcement, contract terms, compliance requirements. Load after the user confirms scope.

Probe the user: "What's the focus — patents/IP, litigation risk, regulatory enforcement, standard contract terms, or compliance in a specific jurisdiction? Any industry or geography?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

⚠️ This skill surfaces publicly available legal information for research purposes. It does not constitute legal advice. Flag when a finding warrants consultation with a qualified lawyer.

## Step 2 axes — IP and patent landscape

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Patent filings** | Key patents in the space (Google Patents, USPTO, EPO): who holds them, filing dates, expiry dates, scope of claims |
| 2 | **Trademark and brand protection** | Registered trademarks (WIPO, USPTO TESS), recent filings, any notable conflicts |
| 3 | **Copyright and licensing** | Content or software licensing norms in the space; open vs proprietary; license compatibility issues |
| 4 | **IP disputes and litigation** | Patent lawsuits, trademark disputes, copyright claims: parties, outcomes, settlements |

## Step 3 axes — Regulatory enforcement and compliance

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Enforcement actions** | Regulatory fines, consent decrees, enforcement notices in the sector (last 5 years): agencies, amounts, stated violations |
| 2 | **Compliance requirements** | What must an operator do to comply? Licenses, certifications, reporting obligations, data handling rules — by jurisdiction |
| 3 | **Upcoming regulatory changes** | Proposed legislation or regulation in consultation, recently passed but not yet in force |

## Step 4 axes — Litigation and contract norms

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Litigation history** | Major lawsuits involving key players: claims, outcomes, appeals, precedents set |
| 2 | **Standard contract terms** | What are the industry-standard terms for key contract types (SaaS MSA, supplier agreements, licensing)? What clauses are contested? |
| 3 | **Jurisdictional risk** | Differences in legal treatment across key jurisdictions (US, EU, UK, APAC); forum-shopping patterns; enforcement asymmetry |

## Citation notes for legal research

Legal sources require specific attribution:

- Court decisions: case name, jurisdiction, year, docket number if available
- Regulations: official register citation (CFR, OJ, etc.) and jurisdiction
- Patent filings: patent number and issuing office
- News about legal matters: verify against official court records when possible; flag "reported" vs "confirmed from filing"

## Section template

```md
## Intellectual Property Landscape

### Key Patents

### Trademarks and Brand Protection

### Copyright and Licensing Norms

### IP Disputes

## Regulatory Enforcement and Compliance

### Enforcement Actions (last 5 years)

### Compliance Requirements by Jurisdiction

### Upcoming Regulatory Changes

## Litigation and Contract Norms

### Litigation History

### Standard Contract Terms

### Jurisdictional Risk

> ⚠️ This report surfaces publicly available legal information for research purposes only. It does not constitute legal advice.
```

## references/market.md

# Market Research — Axes, Probes, and Section Template

Use this file for `type == market` research. Load after the user confirms scope.

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Core market research

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Market size & growth** | TAM/SAM/SOM estimates; CAGR projections; primary data sources (Gartner, IDC, CB Insights, Statista); year of estimate |
| 2 | **Customer segments** | Who buys? Segment breakdown by size, industry, geography; behavioral patterns; willingness to pay signals |
| 3 | **Pricing models** | How do incumbents charge? Per-seat, usage-based, freemium, enterprise license? Price ranges per tier |
| 4 | **Trends & dynamics** | Macro forces driving growth or decline; technology shifts; regulatory tailwinds/headwinds; recent M&A |

For each axis, critical numeric claims (size, growth rate, price) require 2+ sources or `confidence: Low`.

## Step 3 axes — Competitive landscape

Fan out these 4 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Top competitors** | Who are the 5–10 main players? Funding, headcount, founding year, geography |
| 2 | **Positioning & messaging** | How does each player position itself? Target segment, key differentiator, tagline |
| 3 | **Strengths & weaknesses** | What do reviewers (G2, Capterra, Reddit, HN) say? Recurring complaints and praise |
| 4 | **Market gaps** | Underserved segments, feature gaps, pricing gaps, geographic whitespace |

Produce a **comparison table** (competitors × dimensions) at the end of this step's section.

## Step 4 axes — Customer deep dive

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Pain points** | Top 5 problems customers try to solve; evidence from forums, reviews, community posts |
| 2 | **Purchase triggers** | What causes someone to start evaluating? What causes them to choose or reject? |
| 3 | **Customer journey** | Awareness → consideration → decision touchpoints; sales cycle length; champions vs blockers |

## Section template

Use these headings in the output file:

```md
## Market Overview

### Market Size and Growth

### Customer Segments

### Pricing Models

### Trends and Dynamics

## Competitive Landscape

### Key Players

### Positioning Map

### Competitive Matrix

| Player | Segment | Differentiator | Pricing | Strengths | Weaknesses |
| ------ | ------- | -------------- | ------- | --------- | ---------- |

### Market Gaps

## Customer Insights

### Top Pain Points

### Purchase Triggers

### Customer Journey
```

## references/org.md

# Person / Org Research — Axes, Probes, and Section Template

Use this file for `type == person/org` research: due diligence on a company or public figure. Load after the user confirms scope.

Probe the user: "Company or individual? What's the context — investment, partnership, hiring, competitive intelligence, or general background? Any specific concerns to investigate?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

⚠️ Stick to **publicly available information only**. Do not attempt to surface private data, personal contact details, or non-public financial figures.

## Step 2 axes — Core profile

Fan out these 4 axes as sub-agents in a single message:

**For a company:**

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Company background** | Founding story, HQ, headcount, legal structure, subsidiaries, key milestones |
| 2 | **Financial profile** | Funding rounds (Crunchbase, PitchBook), known investors, revenue signals (public filings or press), valuation (if available) |
| 3 | **Leadership** | Founders and C-suite: backgrounds, prior companies, public statements, board composition |
| 4 | **Products and customers** | What do they sell? Known customer logos, case studies, contract announcements |

**For a public figure:**

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Background** | Career history, education, publicly stated biography |
| 2 | **Public output** | Books, papers, talks, interviews, social media presence and themes |
| 3 | **Affiliations** | Board memberships, advisory roles, investments, political donations (if public) |
| 4 | **Reputation signals** | Awards, recognition, criticism, controversies — from reputable sources only |

## Step 3 axes — Press and perception

Fan out these 3 axes (apply to both company and individual):

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Press coverage** | Major news mentions (last 3 years): positive, critical, neutral. Source quality matters — flag tabloids vs tier-1 press |
| 2 | **Controversies and risks** | Lawsuits, regulatory actions, employee complaints (Glassdoor, Blind), public disputes |
| 3 | **Social sentiment** | Twitter/X, LinkedIn, Reddit, HN: community perception, recurring themes |

## Step 4 axes — Strategic context

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Partnerships and ecosystem** | Known partnerships, integrations, joint ventures, industry association memberships |
| 2 | **Hiring signals** | Current job postings and what they reveal about strategic priorities |
| 3 | **Competitive position** | How do they describe themselves vs competitors? What market are they trying to own? |

## Section template

```md
## Profile

### Background

### Financial Profile

### Leadership

### Products and Customers

## Press and Perception

### Press Coverage

### Controversies and Risks

### Social Sentiment

## Strategic Context

### Partnerships and Ecosystem

### Hiring Signals

### Competitive Position
```

## references/parallel-search.md

# Parallel Search — Sub-agent Fan-out

This file describes how to split research work across sub-agents for maximum coverage and speed.

## When to fan out

Fan out whenever a research step has 3+ independent axes that do not depend on each other's findings. Axes are independent when each can be fully researched from scratch without knowing the others' results.

Do not fan out when axes have ordering dependencies (e.g., you need competitor names from Step 3 before you can look up their pricing). In that case, run the dependency first, then fan out the remaining axes.

## Maximum agents per step

Default to **3–5 sub-agents per step**. This range balances coverage, speed, and coordination overhead for most research tasks.

You may exceed 5 — up to **20 in a single step** — only when all three conditions hold: axes are truly independent, the topic is broad enough to justify the extra coverage, and each sub-agent can be given a clear self-contained prompt. Beyond 5, diminishing returns and context overhead increase quickly, so treat larger fan-out as an explicit exception, not the default.

## Sub-agent prompt template

Each sub-agent prompt must be self-contained — the agent has no memory of the parent conversation.

```
You are a research analyst. Your task: research one specific axis of a larger study.

**Topic:** {overall topic}
**Your axis:** {axis name and description}
**Research goals:** {what specific questions to answer on this axis}
**Geographic/time constraints:** {any from scope interview, or "none"}

Instructions:
1. Run web searches and fetch the relevant source pages.
2. For each finding, note the source URL, access date, and confidence level (High/Medium/Low per the ladder below).
3. Tag each source: **Primary** (official docs, government filings, peer-reviewed papers), **Established** (major publications, analyst firms with editorial process), or **Low** (blogs, forums, single opinions). Flag Low-tier sources visibly.
4. Critical claims (numbers, market share, projections) need 2+ sources or get confidence: Low.
5. Flag any conflicts between sources explicitly — do not silently pick one.
6. The axis definition is a starting point, not a ceiling. If you find relevant information that falls outside the stated axis but adds meaningful insight for the overall topic, include it — label it clearly and explain why it matters.
7. External files (PDFs, datasets, analyst reports, regulatory filings, whitepapers, charts) may contain valuable data. When encountered, their key content should be summarized inline — do not leave them as bare links. The `curl` command is available for local downloads when needed.
8. Write findings as **prose paragraphs**, not bullet lists. Embed figures in sentences: "The market reached $4.2B in 2024 [Source]" rather than "* Market: $4.2B". Bullets are acceptable only for true enumerated lists (product names, compliance items, enumerated steps).
9. Distinguish sourced facts from your analysis: use "According to [Source]..." for direct findings and "This suggests..." or "The pattern indicates..." for your synthesis. Never present inference as fact.
10. If a topic cannot be found, write "No sources found for X" — do not guess or leave a blank.
11. Return your findings as a Markdown section ready to paste into a report.

Confidence ladder:
- High: 2+ reputable independent sources agree
- Medium: 1 reputable source (Primary or Established tier)
- Low: blog, forum, single opinion, Low-tier source, or inferred

Citation format: [Source Name](url) (accessed YYYY-MM-DD, confidence: High|Medium|Low)

Output format:
## {Section heading}
{Prose paragraphs with inline citations. Bullets only for true lists.}

> Conflicts noted: {if any}
> Gaps: {what you couldn't find}
```

## Choosing agent type

- Use **`general-purpose`** for broad search + synthesis work (most research axes).
- Use **`Explore`** only when the axis is about exploring an existing local codebase or file set — not for web research.

## Handling completions

Sub-agents complete at different times. As each one finishes:

1. Read its output immediately.
2. Append its findings section to the output file.
3. Do not wait for all agents before writing — write as you receive.

If a sub-agent returns insufficient findings (e.g., a gap note with no data), do not fill the gap with guesses. Write the gap explicitly in the output file and note it for the Synthesis step.

## Splitting axes by type

See the type-specific reference files for the exact axis breakdown:

- `market` → `references/market.md`
- `domain` → `references/domain.md`
- `technical` → `references/technical.md`
- `competitive` → `references/competitive.md`
- `product` → `references/product.md`
- `academic` → `references/academic.md`
- `person/org` → `references/org.md`
- `financial` → `references/financial.md`
- `legal` → `references/legal.md`
- `trend` → `references/trend.md`
- `community` → `references/community.md`
- Other types → no reference file; derive axes from the topic and goals (see Step 1 in SKILL.md)

## references/product.md

# Product Research — Axes, Probes, and Section Template

Use this file for `type == product` research: deep analysis of a specific product. Load after the user confirms scope.

Probe the user: "Which product? Are you a user, potential buyer, or building a competing product? Any specific dimensions (UX, pricing, integrations, reliability, roadmap)?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Product fundamentals

Fan out these 5 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Target audience** | Who is this product built for? Stated personas, industries, company sizes, or roles from the website, docs, and case studies. Who is explicitly excluded or out of scope? |
| 2 | **Feature surface** | What does the product do? Key workflows, supported use cases, notable limitations. Official docs and changelog as primary sources |
| 3 | **Pricing and packaging** | Tiers, price per unit, free vs paid, enterprise pricing signals, recent pricing changes |
| 4 | **Technology and architecture** | What stack does it run on? SaaS vs self-hosted? Performance characteristics, uptime history, known outages |
| 5 | **Ecosystem and integrations** | What does it connect to? API quality, SDK availability, marketplace or plugin ecosystem |

## Step 3 axes — User and market perception

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **User reviews** | G2, Capterra, Product Hunt, App Store, Reddit, HN: recurring praise, recurring complaints, NPS signals |
| 2 | **Changelog and roadmap** | Recent releases (last 12 months): what shipped? Any public roadmap? What's been requested for years and not shipped? |
| 3 | **Community health** | Slack/Discord size, forum activity, GitHub issues response time, docs quality, onboarding experience reports |

## Step 4 axes — Strategic signals

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Competitive positioning** | How does the maker position this product vs alternatives? Who do they name as competitors? |
| 2 | **Adoption signals** | Customer logos, case studies, job postings using this product, conference mentions |
| 3 | **Risk signals** | Recent controversies, pricing hikes, deprecations, reliability incidents, support quality complaints |

## Section template

```md
## Product Overview

### Target Audience

### Feature Surface

### Pricing and Packaging

### Technology and Architecture

### Integrations and Ecosystem

## User and Market Perception

### Review Summary

### Changelog Highlights (last 12 months)

### Community Health

## Strategic Signals

### Competitive Positioning

### Adoption Signals

### Risk Signals
```

## references/technical.md

# Technical Research — Axes, Probes, and Section Template

Use this file for `type == technical` research (technology evaluation, architecture analysis, tool comparison). Load after the user confirms scope.

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

## Step 2 axes — Core technical landscape

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Technology landscape** | What are the main approaches/tools/frameworks in this space? Who are the primary authors or maintainers? |
| 2 | **Performance & benchmarks** | What do published benchmarks show? Latency, throughput, resource usage? What scenarios do benchmarks cover or miss? |
| 3 | **Community & maturity** | GitHub stars/forks/activity, StackOverflow volume, release cadence, CVE history, last major release date |
| 4 | **Integration & compatibility** | What ecosystems does each option fit into? Language support, cloud provider integrations, protocol compatibility |

## Step 3 axes — Comparative analysis

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Top 3–5 options head-to-head** | For each: strengths, weaknesses, license, pricing (open-source vs commercial), vendor lock-in risk |
| 2 | **Decision criteria** | What dimensions matter most for the use case? Performance vs DX vs cost vs operational complexity |
| 3 | **Migration & lock-in** | What does adopting each option commit you to? Migration path if you switch later? Data portability? |

Produce a **decision matrix** (options × criteria) at the end of this step.

## Step 4 axes — Implementation deep dive

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Architecture patterns** | What are the recommended deployment topologies, data models, or integration patterns? What are the trade-offs? |
| 2 | **Common pitfalls** | What gotchas, version quirks, or configuration traps do practitioners report? (GitHub issues, blog post-mortems, HN discussions) |
| 3 | **Security & compliance** | Known CVEs, authentication mechanisms, data-at-rest/in-transit posture, compliance certifications (SOC 2, HIPAA, GDPR) |

## Section template

Use these headings in the output file:

```md
## Technology Landscape

### Overview of Approaches

### Performance and Benchmarks

### Community Health and Maturity

### Integration and Compatibility

## Comparative Analysis

### Top Options Head-to-Head

### Decision Matrix

| Option | Performance | DX  | Community | Cost | Lock-in | Notes |
| ------ | ----------- | --- | --------- | ---- | ------- | ----- |

### Migration and Lock-in Risks

## Implementation Considerations

### Recommended Architecture Patterns

### Common Pitfalls and Gotchas

### Security and Compliance
```

## references/trend.md

# Trend / Foresight Research — Axes, Probes, and Section Template

Use this file for `type == trend` research: emerging signals, weak signals, horizon scanning, future scenario mapping. Load after the user confirms scope.

Probe the user: "What domain or industry? What horizon — 1 year, 3 years, 10 years? Are you looking for early weak signals, near-term bets, or long-range scenarios?"

The axes below are a starting point — include any additional relevant findings encountered during research. Completeness matters more than strict adherence to predefined categories.

**Distinguish clearly between signals (observed, sourced) and forecasts (modeled or opined).** A trend is not a prediction — it is an observed pattern with momentum. Label forecasts as such and cite who made them.

## Step 2 axes — Signal detection

Fan out these 4 axes as sub-agents in a single message:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Analyst and research forecasts** | What do Gartner, Forrester, McKinsey, BCG, IDC, or domain-specific analysts say is coming? Horizon and confidence level per forecast |
| 2 | **Investment signals** | What are VCs, corporates, and governments betting on? Recent funding surges, accelerator focus areas, government grant programs |
| 3 | **Academic and R&D signals** | What research areas are gaining publication volume? What patents are being filed in volume by leading institutions? |
| 4 | **Cultural and behavioral signals** | Search trend data (Google Trends), media coverage spikes, social discourse shifts, consumer behavior changes |

## Step 3 axes — Momentum and adjacent forces

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Technology enablers** | What underlying technology shifts are making this trend possible now vs 5 years ago? What's still missing? |
| 2 | **Adjacent domain spillovers** | What trends in adjacent industries are likely to spill into this domain? What has already crossed over? |
| 3 | **Regulatory and geopolitical tailwinds/headwinds** | What policy changes could accelerate or block this trend? Any country or bloc moving faster than others? |

## Step 4 axes — Scenario mapping

Fan out these 3 axes:

| Agent | Axis | Key questions |
| --- | --- | --- |
| 1 | **Bear case** | What would slow, stall, or reverse this trend? Counterexamples and failed prior predictions in this domain |
| 2 | **Base case** | What's the consensus trajectory? What's the implied timeline? |
| 3 | **Bull case** | What would accelerate this trend beyond current expectations? Historical analogues of faster-than-expected adoption |

## Foresight discipline

Good trend research explicitly distinguishes:

| Label            | Meaning                                               |
| ---------------- | ----------------------------------------------------- |
| **Signal**       | Observed, sourced, happening now                      |
| **Emerging**     | Directional but early; < 5% adoption or awareness     |
| **Accelerating** | Adoption curve past the inflection point              |
| **Forecast**     | Modeled projection with stated assumptions and source |
| **Speculative**  | Plausible but not anchored to current evidence        |

Tag each finding with one of these labels to prevent conflating observation with prediction.

## Section template

```md
## Signal Detection

### Analyst and Research Forecasts

### Investment Signals

### Academic and R&D Signals

### Cultural and Behavioral Signals

## Momentum and Adjacent Forces

### Technology Enablers

### Adjacent Domain Spillovers

### Regulatory and Geopolitical Forces

## Scenario Map

### Bear Case

### Base Case (consensus trajectory)

### Bull Case
```