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

research-verify

Skill49 starsChanged 8 days agoMITView on GitHub

Independently check a research claim against original evidence, reproducible analysis, and plausible alternatives. Use before accepting an important finding or when a result is disputed.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. tony/skills★ 0same name
    2. XiaYiHann/scholar-3★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository6 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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, claimed result, relevant sources or raw artifacts, and planned evaluation criteria.

    line 12
  2. Distinctive:

    Formal reasoning: check premises and critical steps.

    line 22
  3. Distinctive:

    Test the strongest plausible alternative explanation.

    line 23
  4. Distinctive:

    If the project uses a separate completion status, keep it distinct from the scientific verdict.

    line 37

≈ “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

Research Verify

Determine what the evidence warrants. Verification must add an observable check; agreement from another model alone is not independent evidence.

Inputs

Use the research question, claimed result, relevant sources or raw artifacts, and planned evaluation criteria. Request missing raw material only when necessary to assess the claim. Review proportionally to the consequence of error.

Procedure

  1. Restate the claim precisely, including its population, comparison, uncertainty, and scope. Separate descriptive, causal, predictive, and normative claims. Check whether the evidence could support that type of statement.
  2. Inspect original material for each central claim. Verify that source identifiers resolve to the intended material and that the cited section supports the wording. A reference's existence and its relevance are separate checks.
  3. Where possible, assess raw evidence before reading the author's favored interpretation. If another reviewer is available, provide the question, method, and necessary inputs without supplying the desired verdict. Ask for checks and counterexamples, not a confidence vote.
  4. Perform the checks appropriate to the work:
    • Computational: reproduce the main calculation or an independently derived check; inspect sample counts, units, baseline, evaluation split, uncertainty, leakage, and sensitivity to consequential assumptions.
    • Evidence synthesis: inspect source dependence, omitted counterevidence, date and population applicability, inaccessible material, and whether the search supports any claim of completeness.
    • Formal reasoning: check premises and critical steps; use a proof checker when available and appropriate. A proof sketch is not a machine-checked proof.
  5. Test the strongest plausible alternative explanation. Look for a concrete failure case rather than generating generic objections. Preserve disagreements that the available evidence cannot resolve.
  6. Record each check performed and its outcome. Distinguish checks not attempted from checks that failed. If independent review is unavailable, say so and describe the narrower self-check actually completed.
  7. Return a verdict tied to evidence: supported within stated limits, needs revision, contradicted, or unresolved. Recommend the smallest action that would change a consequential unresolved judgment.

Output

Provide a verification note containing:

  • Claim assessed and evidence inspected.
  • Checks actually performed, with artifact or source pointers.
  • Findings that support or weaken the claim.
  • Remaining uncertainty and review limitations.
  • Verdict and any required corrections or next test.

If the project uses a separate completion status, keep it distinct from the scientific verdict. An investigation may be complete while its hypothesis remains unresolved. A successful formatting or schema check does not validate truth.

Completion check

The verdict follows from documented checks, scope is no broader than the evidence, and material objections are resolved or visible. Do not mark a finding verified merely because its author and reviewer agree, or because a script exits successfully.

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.