agentleFS
Codagent-AICodagent-AI/agent-skills/skills/review-assumptions/SKILL.md

review-assumptions

Skill31 starsChanged 6 months agoMITView on GitHub

Reviews risky/notable assumptions and context gaps surfaced by implementor session reports, fixes high-confidence issues directly, and asks one clarifying question at a time for ambiguous findings. Use when the user says "review assumptions", "audit implementor assumptions", or when invoked by a workflow's assumption-review step.

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    Verify claims against the controlling artifacts and relevant source before choosing a disposition:

    line 24
  2. Distinctive:

    Before reporting, reconcile the dispositions with the extracted findings.

    line 35
  3. Distinctive:

    Treat implementor statements as claims, not evidence.

    line 37
  4. Distinctive:

    Run relevant validation and commit applied changes using the project's conventions unless the caller explicitly owns validation or commits.

    line 41
  5. Distinctive:

    Keep the review scoped to reported findings and the source checks needed to evaluate them.

    line 64

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

Review Assumptions

Audit assumptions and context gaps from implementor session reports against the approved plan. Fix clear deviations, ask the user about product ambiguity, and preserve plan-quality gaps for the final summary.

Process

  1. Find the session reports and map each to its workflow iteration and exact task file using available run metadata. Do not infer task identity from report order; label an unmappable report task unknown.
  2. Extract each risky or notable assumption and each context gap. Give every finding a stable identity and ensure none is silently dropped.
  3. Verify claims against the controlling artifacts and relevant source before choosing a disposition:
    • Fix: a clear defect, omission, or deviation from the approved plan. Make the focused change directly, or delegate nontrivial implementation when the caller permits it.
    • Ask: a product, scope, or UX decision that evidence cannot resolve. Use codagent:ask-questions, one finding at a time, with context, practical options, impact, and a recommendation. Apply the user's answer before moving on.
    • Accept: the implementation matches the plan or the stated risk is intentionally acceptable.
    • Defer: the issue is real but outside this review's authority; identify the owner or later decision needed.
    • Context gap: the report explicitly says the plan lacked information. Do not invent a resolution; surface what future planning needed to provide.
  4. Before reporting, reconcile the dispositions with the extracted findings.

Treat implementor statements as claims, not evidence. Inspect the cited code or artifacts before asserting that something is fixed or already handled. Do not turn an ordinary assumption into a context gap merely to avoid resolving it.

Run relevant validation and commit applied changes using the project's conventions unless the caller explicitly owns validation or commits. Do not create an empty commit.

Report

Summarize only applicable sections:

## Review Summary

### Fixed
- [task] — [finding] → [change and commit]

### Resolved with user
- [task] — [finding] → [decision and resulting action]

### Reviewed, no change
- [task] — [finding] → [accepted or deferred, with rationale]

### Context gaps
- [task] — [gap] — Missing: [information the plan needed]

Keep the review scoped to reported findings and the source checks needed to evaluate them. Do not attribute work from later workflow phases to this review.

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.