research

name	research
description	Ethical web research with attribution tracking, license compliance, and citation-based output. Use for technical research, competitive analysis, or content synthesis requiring source transparency.
category	process
triggers	["research topic","find information","web research","cite sources","license check","attribution"]

Research

Purpose: Conduct web research ethically with proper attribution and license compliance. Mode: Read-only output — produces cited research, does not modify project files. Usage: /research [flags] <topic or question>

Iron Laws

1. NEVER PRESENT UNCITED CLAIMS — Every factual assertion must trace to a specific source. If the source is unknown, say "unverified" rather than omitting attribution.

2. RESPECT ACCESS BOUNDARIES — Do not use content from sources that explicitly prohibit AI access (robots.txt disallow, Terms of Service restrictions). If a source is inaccessible, note it and find an alternative.

3. LICENSE DETERMINES USE — Code snippets and creative content carry licenses. Permissive (MIT, Apache, BSD) may be adapted with attribution. Copyleft (GPL, AGPL) requires disclosure. Unknown license means "do not copy verbatim." See references/license-detection.md.

4. FACTS ARE FREE, EXPRESSION IS NOT — Facts, data, and ideas are not copyrightable. The specific wording, structure, and creative expression used to present them are. Always paraphrase and synthesize rather than copying prose.

5. PROVENANCE OVER CONVENIENCE — When two sources conflict, prefer the one with clearer provenance (official docs > blog post). When a source cannot be verified, flag it rather than silently including it.

6. NO MARKET SUBSTITUTION — Research output must not substitute for visiting or purchasing the original source. Summarize and link; do not reproduce.

7. ATTRIBUTE CC-BY CONTENT — Creative Commons Attribution (CC-BY) licensed content requires explicit attribution with author, title, source URL, and license. This is a legal obligation, not optional.

When to Use

• Technical research before implementation decisions
• Technology or library comparison with license analysis
• Finding best practices, standards, or specifications
• Investigating library options with license requirements
• Synthesizing documentation from multiple sources
• Checking license compatibility for a dependency

When NOT to Use

• Codebase exploration → use /explore
• Bug investigation → use /debug
• Security assessment → use /security-review
• Questions answerable from project context alone → use /explore

Scope Flags

Flag	Description	Default
--depth=<quick|standard|deep>	Research depth	standard
--format=<summary|report|comparison>	Output format	summary
--license-check	Focus on license analysis for specific libraries/content	off

Content Classification

Category	Copyright Status	Handling
Facts and data	Not copyrightable	Freely usable; cite for credibility
Standards and specs	Varies (often CC-BY)	Check license; attribute if required
Code snippets	License-dependent	Classify per references/license-detection.md
Creative expression	Protected	Paraphrase and synthesize; never copy verbatim
Public domain / CC0	Not protected	Freely usable; cite for provenance

Confidence Levels

Level	Criteria	Handling
HIGH	Primary source, official documentation, peer-reviewed	Present as established fact
MEDIUM	Reputable secondary source, well-maintained blog, verified community answer	Present with source qualification
LOW	Single unverified source, personal blog, outdated content	Flag as unverified; seek corroboration

Workflow

Phase 1: Scope

1. Parse the research question — what specifically needs answering?
2. Classify the research type:
   • Factual lookup — quick answer from authoritative source
   • Comparison — structured evaluation of alternatives
   • Deep research — comprehensive analysis with multiple perspectives
   • License check — focused license and compliance analysis
3. Select depth level based on --depth flag or question complexity
4. Identify relevant source types: official docs, specs, RFCs, code repos, articles

Phase 2: Gather

For each source accessed:

1. Record provenance:
   • URL, title, author/organization, publication date
   • License (if stated — check footer, LICENSE file, meta tags)
   • Access method and any restrictions observed
2. Classify content type per the Content Classification table above
3. Check access boundaries:
   • Was this content behind a paywall or login? → Do not use
   • Does the site's ToS prohibit AI use? → Note and find alternative
4. For code: identify license using the detection order in references/license-detection.md
5. Cross-reference claims across multiple sources when possible

Phase 3: Analyze

1. Synthesize — paraphrase and combine findings; never copy prose verbatim
2. Resolve conflicts using source hierarchy:
   • Primary/official documentation
   • Reputable secondary sources
   • Community sources (Stack Overflow, blog posts)
3. License classification for any code to be included:
   • Run through references/license-detection.md classification
   • Determine obligations: attribution, notice, share-alike, or prohibited
4. Assign confidence levels to each finding

Phase 4: Cite and Verify

Run the compliance checklist from references/compliance-checklist.md:

• Every factual claim has a source
• No verbatim copying of creative expression
• Code snippets have license classification
• CC-BY content has proper attribution (see references/citation-formats.md)
• Copyleft code flagged with obligations, not silently included
• Source access boundaries were respected
• Output does not substitute for visiting original sources

Phase 5: Output

Format based on --format flag:

Summary (default)

## [Topic]

[Synthesized findings with inline citations [1][2]...]

### Sources

| # | Source | Author/Org | Date | License | Confidence |
|---|--------|------------|------|---------|------------|
| [1] | [Title](URL) | Author | Date | License | HIGH/MED/LOW |

Report

## [Topic]

### Background
[Context and scope]

### Findings
[Structured sections with inline citations]

### Analysis
[Synthesis, trade-offs, recommendations]

### Sources
[Full source table as above]

### License Obligations
[Only if code or CC-BY content included — what the consumer must do]

Comparison

## [Topic]: Comparison

| Criterion | Option A | Option B | Option C |
|-----------|----------|----------|----------|
| [aspect] | [value] [1] | [value] [2] | [value] [3] |

### Recommendation
[Based on findings]

### Sources
[Full source table]

All formats end with a Sources table and, when applicable, a License Obligations section listing what the consumer must do with any included licensed content.

Quick Reference

Task	Command
Quick lookup	/research --depth=quick what is X
Compare technologies	/research --format=comparison React vs Vue vs Svelte
Check a library's license	/research --license-check lodash
Deep technical report	/research --depth=deep --format=report state of WebAssembly

Acceptance Tests

ID	Type	Prompt / Condition	Expected
RES-T1	Positive	"Research best practices for caching"	Skill triggers
RES-T2	Positive	"Find information about WebSocket libraries"	Skill triggers
RES-T3	Positive	"What license does lodash use"	Skill triggers
RES-T4	Positive	"Compare X vs Y"	Skill triggers
RES-T5	Negative	"How does our auth module work?"	Does NOT trigger → /explore
RES-T6	Negative	"Fix this bug"	Does NOT trigger → /debug
RES-T7	Boundary	"Research and then implement X"	Triggers for research phase only