agentleFS
Sign inStart free
bordenetbordenet/superpowers-plus/skills/engineering/micro-harsh-review/skill.md

micro-harsh-review

Skill8 starsChanged 6 months agoMITView on GitHub

Per-batch adversarial review for ANY code change. 3 critic personas score on 5 dimensions each. Score <8 average = REJECT + rework. Faster than full review, more rigorous than lint.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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:

    Before committing ANY code change (even 1 function)

    line 40
  2. Distinctive:

    After modifying logic, not just formatting

    line 41
  3. Distinctive:

    Before pushing to shared branch

    line 42
  4. Distinctive:

    When editing router patterns or gate thresholds

    line 43
  5. Distinctive:

    Full access: START FROM the local diff.

    line 60
  6. Distinctive:

    Full access: START FROM interface contracts and public APIs.

    line 72
  7. Distinctive:

    Full access: START FROM failure modes and state transitions.

    line 78

≈ “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

Micro Harsh Review

Wrong skill? Full PR -> progressive-code-review-gate. Non-code -> progressive-harsh-review. Style -> enforce-style-guide.

Announce at start: "I am using the micro-harsh-review skill to review this change."

When to Use

  • Before committing ANY code change (even 1 function)
  • After modifying logic, not just formatting
  • Before pushing to shared branch
  • When editing router patterns or gate thresholds

Example

# Example: 3-critic scoring
echo "=== Micro Harsh Review: router pattern change ==="
echo "Critic 1 (Nitpick):  8/10 - case sensitivity concern"
echo "Critic 2 (Arch):     9/10 - minimal scope, pattern-local"
echo "Critic 3 (Prod):     7/10 - no collision test added"
echo "Average: 8.0 -> PASS (conditional: add collision test)"

3-Critic Protocol

Critic 1: NitpickLineByLine

Full access. START FROM the local diff. PRIORITIZE line-by-line syntax, naming, and surface-level correctness.

CheckScore /10Notes
Off-by-one
Null handling
String comparison
Error messages
Variable naming

Critic 2: ArchSoundnessProbe

Full access. START FROM interface contracts and public APIs. PRIORITIZE ripple analysis across callers and downstream consumers.

Respects patterns? /10 . Downstream impact? /10 . Minimal scope? /10 . 10x load? /10 . Reversible? /10

Critic 3: ProdBattleTest

Full access. START FROM failure modes and state transitions. PRIORITIZE error handling, retry/rollback logic, and 3 AM operational safety.

Edge cases? /10 . Failure mode? /10 . Logging? /10 . Backward compat? /10 . Ship at 3 AM? /10

Scoring

Average = (Nitpick + Arch + Prod) / 3

  • = 8.0 -> PASS

  • = 6.0 -> CONDITIONAL — fix flagged, re-review

  • < 6.0 -> REJECT -> think-twice -> redo

Anti-Patterns

Anti-PatternDetectionCorrection
Rubber-stampAll scores 9-10, no notesFind >=1 concern per critic
Inflated scoresAverage > 9.0 consistentlyRecalibrate with known-bad code
Critics agreeSame findings across all 3Force second-order critique: each must name ≥1 failure mode or cite a specific property of the change explaining why none exists (generic "it's straightforward" = rubber-stamp)
Review > coding time>15 min for <20 linesTop 3 risks only

Failure Modes

FailureDetectionRecovery
Rubber-stamp (all 10s)No notesFind >=1 concern per critic
Review > change time>15 min for <20 linesTop 3 risks only
Critics agree on everythingSame findingsForce second-order critique: name ≥1 plausible failure mode or cite specific change property explaining why none exists
Score inflationAverage > 9.0 consistentlyRecalibrate with bad code

Companion Skills

  • progressive-code-review-gate: Full PR review (heavier)
  • pre-commit-gate: Pre-commit (lighter, automated)
  • enforce-style-guide: Style enforcement
  • think-twice: Deeper problem revealed
  • failure-autopsy: Reviewed code fails later

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.

Check your setup, free