agentleFS
yagizdoyagizdo/quiver/skills/report-check/SKILL.md

report-check

Skill16 starsChanged 57 days agoMITView on GitHub

Analyze a review report for quality -- detects noise, false positives, overkill suggestions, and findings that exist to appear thorough. Usage: /report-check <path-to-report>

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 13 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If $ARGUMENTS is empty, print:

    line 29
  2. Very specific:

    If $ARGUMENTS contains a file path:

    line 36
  3. Distinctive:

    Read the file using the Read tool.

    line 37
  4. Distinctive:

    If the file does not exist or is not readable, print:

    line 38
  5. Distinctive:

    Read the full report file.

    line 46
  6. Distinctive:

    If git is available AND a branch name was extracted from the report:

    line 54
  7. Distinctive:

    Use main if it exists, otherwise master.

    line 60
  8. Distinctive:

    If the diff is non-empty, it will be passed to the agent for cross-referencing.

    line 67
  9. Distinctive:

    If git is unavailable or no branch info exists, proceed without diff.

    line 73
  10. Distinctive:

    If no diff is available, include:

    line 94
  11. Distinctive:

    If agent returns zero issues:

    line 107
  12. Distinctive:

    If agent returns issues, display the structured findings to the user.

    line 110
  13. Distinctive:

    After applying all fixes, recalculate:

    line 127
  14. Distinctive:

    Write the updated report back to the same file path.

    line 133
  15. Distinctive:

    Read the modified report file back using the Read tool.

    line 149

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Gather Context

!`git rev-parse --is-inside-work-tree 2>/dev/null || echo "NO_GIT"`
!`git branch --show-current 2>/dev/null || echo "NO_GIT"`

Report Quality Check

Analyze a review report for quality issues -- noise, false positives, overkill suggestions, and findings that exist to appear thorough rather than to help the developer.

Announce: "Using the report-check skill to audit report quality."


Step 0 -- Validate Input

If $ARGUMENTS is empty, print:

Usage: /report-check <path-to-review-report.md> Analyzes a review report for quality issues -- noise, false positives, overkill suggestions.

Stop here.

If $ARGUMENTS contains a file path:

  1. Read the file using the Read tool.
  2. If the file does not exist or is not readable, print:

    Report file not found: {path} Stop here.


Step 1 -- Read Report

  1. Read the full report file.
  2. Extract the branch name from the ## Review Context section if present (look for a Branch: line).
  3. Note the report structure: findings grouped by severity, verdict, filtered findings section.

Step 2 -- Reconstruct Diff (optional)

If git is available AND a branch name was extracted from the report:

  1. Determine the base branch:
    !`git rev-parse --verify main 2>/dev/null || echo "NO_MAIN"`
    Use main if it exists, otherwise master.
  2. Get the diff:
    !`git diff {base}...{branch} 2>/dev/null || echo "NO_DIFF"`
  3. If the diff is non-empty, it will be passed to the agent for cross-referencing.

Print status:

  • With diff: Analyzing report with diff cross-reference...
  • Without diff: Analyzing report (no diff available for cross-reference)...

If git is unavailable or no branch info exists, proceed without diff. The agent works report-only -- Phase 2 accuracy checks will be limited.


Step 3 -- Dispatch Agent

Spawn the quiver:report-checker agent with a self-contained prompt containing:

  1. The full report content, clearly delimited:
    <report>
    {full report markdown}
    </report>
  2. The diff (if available), clearly delimited:
    <diff>
    {full diff output}
    </diff>
    If no diff is available, include:
    <diff>
    No diff available for cross-referencing.
    </diff>
  3. Instruction: "Audit this review report for quality. Apply all three phases (substance verification, accuracy check, proportionality audit). Return structured findings using your output format."

Step 4 -- Present Results

If agent returns zero issues:

Report passed quality audit -- no issues found.

If agent returns issues, display the structured findings to the user.

Use AskUserQuestion:

Quality audit found {N} issues in the report. What would you like to do?

Buttons:

  • Apply fixes to the report
  • Show details only
  • Skip -- keep as-is

Handle each option:

<!-- SYNC: The apply-fixes procedure below (REMOVE/DOWNGRADE/REWRITE actions + recalculation steps) is duplicated in skills/review/SKILL.md Step 3.5 "Handle results" block. Keep both in sync. -->

Apply fixes:

  1. For each issue with action REMOVE: delete the finding from the report file.
  2. For each issue with action DOWNGRADE: change the finding's severity and move it to the correct severity section.
  3. For each issue with action REWRITE: replace the finding's recommendation text with the corrected version. It changes only the recommendation text, so the finding keeps its Verification: line.
  4. After applying all fixes, recalculate:
    • Findings overview counts in ## Review Context
    • Disposition counts in ## Review Context, and the **Deferred:** line under the fix order table
    • Severity section contents (move downgraded findings, remove deleted ones)
    • Recommended Fix Order table (remove entries for deleted/downgraded findings)
    • Verdict line (recompute based on remaining finding severities)
  5. Write the updated report back to the same file path.
  6. Proceed to Step 5.

Show details:

  1. Print the full audit output.
  2. Re-ask via AskUserQuestion:

    Apply the suggested fixes? Buttons: Apply fixes / Skip -- keep as-is

Skip: Stop. Report unchanged.


Step 5 -- Verify (if fixes applied)

  1. Read the modified report file back using the Read tool.
  2. Confirm no {placeholder} text remains in the report.
  3. Print:

    Report updated: {N} findings removed, {M} downgraded, {K} rewritten.

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.