Content
81%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.
A dense, highly actionable offensive-security skill with executable recon and exploit patterns and a clear safe-pattern/detection section. The main gap is progressive disclosure — everything is packed into one file with no supporting bundle files — and the exploit workflow could use an explicit pre-exfil validation checkpoint.
Suggestions
Split the Detection signatures and Tools tables into a references/ file (e.g. references/detection.md) and link to it from SKILL.md so the main file stays a tight overview.
Add an explicit numbered validation step before the exfil-truncation step in the Decision gate (e.g. '4. Confirm only the first 8-12 chars of the token are exfiltrated; abort if more would leak').
Consider extracting the full exploit-payload worked example into a references/exploits.md and keeping a one-line summary inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean, table-and-code-driven layout that assumes Claude's competence and avoids explaining basic Actions concepts; a few sections (e.g. the multi-path explanation under pull_request_target) could be tightened, but no padding. | 4 / 5 |
Actionability | Copy-paste-ready grep recon commands, a fully worked vulnerable workflow plus the matching expanded shell exploit, branch-name injection commands, scope-check curl/gh calls, and SHA-vs-tag pinning examples — concrete and executable throughout. | 5 / 5 |
Workflow Clarity | Clear sequence from recon → sink identification → exploit → safe pattern → detection, plus an explicit Decision Gate with validation/cleanup steps; capped below 5 because the destructive exploit path lacks a hard validation checkpoint before exfil-truncation and the recon→exploit handoff is implied rather than a numbered checklist. | 4 / 5 |
Progressive Disclosure | Well-organized sections with a References list, but all substantive content is inlined in the single SKILL.md with no bundle files (references/, scripts/, assets/ absent) and only one outward link (to poisoned-pipeline-execution/SKILL.md); the detection/tools tables could live in separate reference files. | 3 / 5 |
Total | 16 / 20 Passed |