agentsclimarketplace

Skill review

Skill fabioc-aloha/Alex_Skill_Mall/plugins/supervisor-fleet/skill-review

Review submissions to Alex_ACT_Edition for ACT spec compliance, content quality, and scope fitFrom its SKILL.md

Install
npx -y skills add fabioc-aloha/Alex_Skill_Mall --skill skill-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.8 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

Skill Review

Evaluate a candidate skill, instruction, or prompt for inclusion in Alex_ACT_Edition (the heir template). The Supervisor accepts only what passes all four gates below.

When to Use

  • A PR opens against Alex_ACT_Edition proposing a new skill/instruction/prompt
  • A .skill.review.md file lands in Alex_ACT_Supervisor/feedback/inbox/
  • A heir's curation log proposes promoting a custom skill to the Edition template
  • Triage routes a feedback item to "Edition fix"

The Four Gates

A submission must pass all four to be merged. Failure on any gate = rejection or revision.

Gate 1 — Spec Compliance

CheckPass criterion
Frontmatter presentYAML block with type, lifecycle, inheritance, description, application, applyTo, currency
applyTo glob is non-emptyAt least one path pattern, no **/* unless explicitly justified
inheritance set correctlyinheritable for fleet-wide; custom for heir-only; master-only reserved for Supervisor/AlexMaster
File location matches type*.instructions.md in instructions/, SKILL.md in skills/<name>/, *.prompt.md in prompts/
Markdown lints cleanNo broken links, no missing code-fence languages

Gate 2 — Content Quality

CheckPass criterion
Single responsibilityThe artifact does one thing; if title contains "and" or "+", split it
Behavioral, not encyclopedicTells the agent what to do, not what a topic is
Has falsifiability or visible markersThe reader can tell whether the artifact fired correctly
≤ 300 lines (skills) / ≤ 200 lines (instructions)Longer = signal of overload; split or trim
No duplicated content from existing artifactsGrep for overlapping descriptions across .github/instructions/

Gate 3 — Scope Fit

CheckPass criterion
ACT-Edition targetBelongs in the heir template, not in AlexMaster or a different edition
Not framework-levelDoes not modify ACT manifesto, tenets, or claims registry — that's AlexMaster's territory
Not a one-off project scriptA skill must generalize to ≥2 projects; if only one heir would ever use it, it's custom, not inheritable
Doesn't duplicate Skill Mall contentIf the value is a marketplace listing, it goes in Alex_Skill_Mall, not the Edition

Gate 4 — Safety

CheckPass criterion
No destructive defaultsAnything that deletes, force-pushes, or overwrites must require explicit user approval
No hardcoded credentials or PIIRun pii-memory-filter mentally over the diff
No prompt-injection vectorsIf the artifact reads external content (URLs, files), it sanitizes or quotes it
ReversibleA user can disable or remove the artifact without breaking the Edition

Decision Matrix

Gates passedAction
All 4Accept — merge and add to next Edition release
3 of 4Request revision — leave a review comment naming the failed gate
2 of 4 or fewerReject — close the PR with a written rationale; add to decisions/ if the rejection sets precedent

Output Template

When the review completes, write a verdict to feedback/inbox/<submission-id>.review.md:

# Review: <submission title>

**Submission**: <path or PR#>
**Reviewer**: Supervisor
**Date**: <ISO date>
**Verdict**: Accept | Request revision | Reject

## Gate Results

- Gate 1 (Spec): Pass / Fail — <details>
- Gate 2 (Quality): Pass / Fail — <details>
- Gate 3 (Scope): Pass / Fail — <details>
- Gate 4 (Safety): Pass / Fail — <details>

## Rationale

<2-3 sentence summary of the verdict>

## Required Changes (if revision)

- <bullet list of concrete asks>

## ACT Pass Trail

<paste the trimmed act-pass markers from the review>

Anti-Patterns

Anti-patternCorrection
Approving because the submitter is confidentConfidence ≠ quality. Run all four gates regardless of submitter
Rejecting without naming the gateAlways cite the specific gate; vague rejections waste cycles
Accepting "trivial" submissions without reviewTrivial-looking submissions are where regressions hide
Reviewing without an act-pass trailThis is medium-stakes work — the trimmed pass is mandatory

Related

Keep looking

Skills are one crate of 326,569. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.