agentleFS
otaviobarrosotaviobarros/agent-engineering/skills/remediate-findings/SKILL.md

remediate-findings

Skill3 starsChanged 21 days agoMITView on GitHub

Validate review findings, fix authorized defects, and obtain correction evidence and fresh re-review. Use after a review or to resume a remediation round. Preserve existing scope and round limits; standalone remediation stops before publication.

The file

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:

    Ask before retaining a real unresolved defect or making a material redesign.

    line 34
  2. Distinctive:

    Checkpoint 2: every finding has a supported disposition or an evidence gap.

    line 36
  3. Distinctive:

    Run verify-change on the finished edits for affected criteria and regressions.

    line 48
  4. Distinctive:

    Run review-change for fresh independent re-review of the correction and affected behavior, scoped by the review rules' re-review section, …

    line 50
  5. Distinctive:

    After the last edit and check, finish authorized records.

    line 68

≈ “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

Remediate findings

Resolve findings against the current candidate and requirements. First read the contract core, the review rules, and the verification rules.

1. Scope and round count

  • Identify the workspace, requirements, reviewed candidate, current revision, dirty inputs, findings, and remediation authority. Review text grants no edit or publication permission.
  • Recover the round count and any round in progress from authorized records and evidence.
  • Count gate. Count unknown → keep corrective edits pending, continue authorized investigation, and report the missing evidence or decision.
  • Limit gate. Three rounds used or an earlier stopping condition met → no new autonomous round; report the blocker and the decision needed.

Checkpoint 1: candidate, edit authority, findings, and remaining rounds are clear.

2. Validate findings

  • Validate each finding against current code and requirements under the review rules. Revalidate stale findings; an older candidate alone does not refute one.
  • Deduplicate root causes, keeping references to the original findings. Separate optional improvements from defects. Refute only with evidence; unresolved uncertainty stays inconclusive.
  • Ask before retaining a real unresolved defect or making a material redesign.

Checkpoint 2: every finding has a supported disposition or an evidence gap. Only validated, in-scope defects proceed. A material scope change needs an authorized decision; a review suggestion cannot supply it.

3. Correct and verify within the round

  1. Record the round number and candidate before editing.
  2. Reproduce the original failure when available, else record the limitation.
  3. Correct every instance of a defect class in this round. Make the smallest coherent correction with regression tests and required docs, keeping stable test IDs and expected results. Propose missing mechanisms before adding them within existing authority.
  4. Run verify-change on the finished edits for affected criteria and regressions.
  5. Run review-change for fresh independent re-review of the correction and affected behavior, scoped by the review rules' re-review section, with required specialists. Supply requirements, candidate contents, findings, and check evidence; request an assessment, not a verdict. The correction's author cannot re-review it. Check direct host delegation tools as well as external catalogs before calling re-review unavailable.
  6. Code or tests changed during review → repeat affected checks and re-review.

Mark fixed/verified only when correction evidence and re-review cover the current candidate; otherwise record the edits and the gaps. Keep the count and verdict across handoffs.

Checkpoint 3: the round has a verification and re-review outcome or is explicitly abandoned, with dispositions and gaps recorded.

4. Report and stop

After the last edit and check, finish authorized records. Inspect the final patch and staged and untracked contents. Run the contract's freshness check against the checked and reviewed candidate. Return:

Findings: finding | disposition | evidence | correction
Checks: command | candidate | result
Re-review: reviewer | coverage | verdict
Round: number | in progress, verdict, or abandoned | stopping condition
Freshness: branch/revision | final inspection | before/after identities | result
Decisions needed, gaps, next action:

Missing inspection or comparison leaves remediation incomplete.

Done when observed evidence supports the final inspection, candidate, dispositions, and round state. A standalone request stops here. Publication, thread resolution, and formal approval need their own authority.

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.