Content
83%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.
An excellent, lean, fully-executable attack playbook with strong actionability and conciseness. Its main gap is the absence of validation checkpoints for a destructive workflow, which caps workflow_clarity; adding a verify-escape step would lift it.
Suggestions
Add an explicit validation checkpoint after the exploit, e.g. 'Verify host root: `chroot /host id` should show uid=0' before proceeding to persistence.
Split Persistence/OPSEC/Detection into a separate reference file to bring SKILL.md under ~50 lines and let progressive_disclosure reach 5.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is almost entirely executable commands with no concept explanation (no 'what Docker is' padding); the one-line intro ('fully equivalent to root on the host') earns its place, matching the 'lean and efficient; every token earns its place' anchor. | 5 / 5 |
Actionability | Copy-paste-ready commands for Detect, Docker/containerd/CRI-O exploit paths, a curl-on-UNIX-socket API fallback, persistence, and OPSEC — concrete examples cover the common cases. | 5 / 5 |
Workflow Clarity | Sections flow Detect → Exploit → Persistence → OPSEC, but this is a destructive skill (host RCE, backdoor/SUID creation) with no validation/verification checkpoints (e.g. 'confirm host root via id on /host'); the destructive-ops cap holds workflow_clarity at 3. | 3 / 5 |
Progressive Disclosure | Self-contained, well-organized with clear section headers and no nested references, but at ~80 lines it exceeds the 'under 50 lines' simple-skill threshold that would allow a clean 5, so it sits at 'good structure; minor organization gaps'. | 4 / 5 |
Total | 17 / 20 Passed |