Content
71%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with executable commands, explicit validation, and clear safety framing, which is its core strength. Its weaknesses are structural: a large inline verification script that belongs in a separate file, a dangling reference to a non-existent hooks/ path, and some redundancy between Operational Notes and the frontmatter.
Suggestions
Move the Release Artifact Verification bash block into scripts/ (e.g. verify_release.mjs or verify_release.sh) and keep a one-line invocation plus the pubkey fingerprint in SKILL.md, reclaiming ~70 lines of context.
Fix or remove the reference to hooks/clawsec-advisory-guardian/lib/reputation.mjs — the file does not exist in the bundle, so the manual wiring instructions cannot be followed as written.
Consolidate the three installation paths (Vercel Skills CLI, clawhub install, direct enhanced_guarded_install.mjs) into one section with a short decision note, and drop the Operational Notes lines that duplicate frontmatter requires/depends_on.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and operational with no tutorials on concepts Claude already knows, but the ~70-line Release Artifact Verification bash script is inlined wholesale in SKILL.md and 'Operational Notes' restates the frontmatter's requires/depends_on, so it could be tightened. | 3 / 5 |
Actionability | Fully executable, copy-paste-ready guidance throughout: exact install commands, a complete signed-manifest verification script with set -euo pipefail, fingerprint checks, and error exits, concrete usage invocations with flags, an exit-code table, and a documented environment variable with its default. | 5 / 5 |
Workflow Clarity | There is a clear sequence with explicit validation checkpoints (fingerprint mismatch exits 1, 'Only install or extract the archive after this verification succeeds') and a feedback loop (below-threshold score requires rerunning with --confirm-reputation), but the install flow is fragmented across three alternative paths and the advisory-hook wiring is left manual, leaving minor gaps versus the level-5 anchor. | 4 / 5 |
Progressive Disclosure | Sections are well organized and all three scripts/ references resolve to real bundle files, but the 70-line verification block is inline content that belongs in a separate script or reference file, and the body references hooks/clawsec-advisory-guardian/lib/reputation.mjs which does not exist in the bundle. | 3 / 5 |
Total | 15 / 20 Passed |