agentleFS
laststancelaststance/skills/skills/source-grounded-research/SKILL.md

source-grounded-research

Skill1 starsChanged 56 days agoMITView on GitHub

Cited research briefs

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

24 rules. About 0 are common, 18 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈18

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill for research-only tasks that should end in a cited brief rather than code changes.

    line 11
  2. Distinctive:

    Do not use this skill for:

    line 22
  3. Distinctive:

    Start from the lane's highest-authority source.

    line 31
  4. Distinctive:

    Check for contradictions or recent changes.

    line 33
  5. Very specific:

    Start with official docs via Context7 when available.

    line 40
  6. Very specific:

    Use get_code_context_exa for examples and practical usage.

    line 41
  7. Very specific:

    Use Perplexity or WebSearch only for contradiction checks or recent changes.

    line 42
  8. Very specific:

    Start with DeepWiki when available.

    line 47
  9. Distinctive:

    Read repo structure before asking detailed questions.

    line 48
  10. Very specific:

    Start with Perplexity Search or standard WebSearch.

    line 53
  11. Distinctive:

    Verify final claims with official release notes, changelogs, or docs.

    line 54
  12. Distinctive:

    Start with local repo evidence tools first.

    line 58
  13. Distinctive:

    Start with breadth discovery.

    line 63
  14. Distinctive:

    Run an authority pass.

    line 64
  15. Distinctive:

    Run a contradiction pass before recommending.

    line 65
  16. Distinctive:

    Prefer tools by role, not habit:

    line 69
  17. Distinctive:

    Every material factual claim must have an inline citation.

    line 95
  18. Distinctive:

    Prefer direct URLs to primary sources.

    line 96
  19. Distinctive:

    If only secondary evidence exists, say so explicitly.

    line 97
  20. Distinctive:

    Keep facts and inference separate.

    line 98
  21. Distinctive:

    If a tool is unavailable, switch to another tool with the same role.

    line 102
  22. Very specific:

    If evidence is thin, say insufficient evidence.

    line 103
  23. Distinctive:

    If there is no primary source, lower confidence and state the limitation.

    line 104
  24. Distinctive:

    If evidence conflicts, preserve both sides with citations.

    line 105

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Source-Grounded Research

Purpose

Use this skill for research-only tasks that should end in a cited brief rather than code changes.

This skill is for:

  • web research
  • library and framework documentation lookup
  • open-source repository investigation
  • comparison and decision support
  • recent changes and release-note checks

Do not use this skill for:

  • implementation work
  • code editing
  • speculative answers without sources

Quick Start

  1. Restate the question and classify it into one primary lane.
  2. Start from the lane's highest-authority source.
  3. Corroborate major claims with a second source.
  4. Check for contradictions or recent changes.
  5. Produce a cited report with a query log and sources section.

Research Lanes

library-framework

  • Start with official docs via Context7 when available.
  • Use get_code_context_exa for examples and practical usage.
  • Use Perplexity or WebSearch only for contradiction checks or recent changes.

oss-repository

  • Start with DeepWiki when available.
  • Read repo structure before asking detailed questions.
  • Corroborate key claims with GitHub docs, repo pages, Exa, or official docs.

recent-changes

  • Start with Perplexity Search or standard WebSearch.
  • Verify final claims with official release notes, changelogs, or docs.

local-plus-external

  • Start with local repo evidence tools first.
  • Bridge to external docs only after the local behavior is understood.

comparison-decision

  • Start with breadth discovery.
  • Run an authority pass.
  • Run a contradiction pass before recommending.

Tool Priority

Prefer tools by role, not habit:

  • Sequential Thinking: scope, sub-questions, assumptions
  • Context7: official library/framework docs
  • DeepWiki: OSS repository architecture and topic map
  • Exa: breadth discovery and code/doc context
  • Perplexity: recent web-grounded synthesis and contradiction checks
  • WebSearch / WebFetch: standard web fallback or direct page retrieval
  • local repo tools: first-class evidence when the question touches local code

Output Contract

Always include:

  1. Executive Summary
  2. Evidence Map
  3. What We Know
  4. What We Infer
  5. Contradictions and Unknowns
  6. Confidence
  7. Actionable Next Steps
  8. Query Log
  9. Sources

Citation Rules

  • Every material factual claim must have an inline citation.
  • Prefer direct URLs to primary sources.
  • If only secondary evidence exists, say so explicitly.
  • Keep facts and inference separate.

Failure Rules

  • If a tool is unavailable, switch to another tool with the same role.
  • If evidence is thin, say insufficient evidence.
  • If there is no primary source, lower confidence and state the limitation.
  • If evidence conflicts, preserve both sides with citations.

Report Template

Use this structure:

## Executive Summary
- Key takeaway with citation

## Evidence Map
- Primary: why this source is authoritative
- Supporting: what it adds
- Counter: what challenges the main claim

## What We Know
- Fact with citation

## What We Infer
- Inference linked back to evidence

## Contradictions and Unknowns
- Open disagreement or missing source

## Confidence
- high | medium | low

## Actionable Next Steps
- Specific follow-up

## Query Log
| Tool | Query/Prompt | Filters | Why Used | Outcome |
| --- | --- | --- | --- | --- |

## Sources
- [Title](https://example.com) - Publisher, date

Additional Guidance

  • For the detailed lane matrix and DeepWiki workflow, see reference.md.
  • For concrete example prompts and output shapes, see examples.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.