agentleFS
wrencanflywrencanfly/research-loop-skills/skills/research-synthesize/SKILL.md

research-synthesize

Skill49 starsChanged 8 days agoMITView on GitHub

Turn checked research evidence into a clear answer, reproducibility handoff, and audience-appropriate communication. Use to close an investigation or report a bounded interim result.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the research question, evidence map, results, verification findings, and output requirements.

    line 12
  2. Distinctive:

    Use the requested artifact.

    line 28
  3. Distinctive:

    Do not force this structure onto a short answer that already conveys the essentials.

    line 35

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “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

Research Synthesize

Answer the user's question with the strongest statement the evidence supports. Match the user's language, audience, and requested format.

Inputs

Use the research question, evidence map, results, verification findings, and output requirements. If review is missing, perform a proportionate source and reasoning check or identify the result as unverified. Do not imply independent review that did not occur.

Procedure

  1. Lead with the answer and its scope. If evidence is insufficient or conflicting, make that the answer rather than forcing a definitive conclusion.
  2. Distinguish findings from interpretation and proposed action. Attribute source observations; label the framework's own design choices and extrapolations as proposals. Never convert a scenario, expectation, or association into an observed causal result.
  3. Link consequential claims to inspected sources or produced artifacts. Preserve material caveats, negative findings, and alternative explanations. Describe confidence through evidence and limitations; avoid unsupported numerical certainty.
  4. Explain what changed the initial view and what remains unresolved when this helps the reader assess the result. Separate an absence of evidence from evidence against a claim.
  5. Give a practical handoff proportional to the task: key sources, data and code references, method, reproducible entry point, and necessary dependencies. Distinguish proposed work from executed work and simulated data from observations.
  6. State why this investigation stops now: its evidence standard was met, its budget ended, a bottleneck requires another input, or additional work would add little information. Identify the next action most likely to change the answer without initiating unrequested work.
  7. Adapt for communication only after the claims are stable. In a short post, retain the qualification that changes the meaning; omit less important details instead. Verify publication status before saying a project is released or a result is independently reproduced.

Drafting does not authorize sending, posting, or submitting. Carry out publication only when the user's existing instructions authorize it; preserve the intended destination, audience, and scope.

Output

Use the requested artifact. A typical research handoff includes:

  • Answer and consequential evidence.
  • Interpretation, limitations, and unresolved questions.
  • Reproduction or source-inspection entry point.
  • Recommended next step and stopping reason.

Do not force this structure onto a short answer that already conveys the essentials. If a publication draft is requested, provide usable copy and keep editorial notes outside it.

Completion check

A reader can distinguish what is known, inferred, proposed, and still uncertain. Every public-facing claim remains consistent with the evidence and actual work completed. The output answers the original question at the level the research can sustain.

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.