| name | content-aware-audit |
| description | Authorized deep vulnerability research with CVSS, evidence, repro, impact, and fixes. Use for security audit, exploit chains, authz/injection review — replaces disabled OmO security-review skill for depth. |
| compatibility | opencode |
| license | MIT |
| metadata | {"openconfig":"1.5.30","replaces":"security-review"} |
content-aware-audit
Deep authorized vuln research. Prefer this over re-enabling OmO security-* skills (they hang headless oc run via skills.urls).
When to use
- User asks for security audit, vuln review, exploit path, CVSS, hard authz/injection analysis
- After
content-aware-recon / content-aware-fast surface map
content-aware-deep category / content-aware-research agent / content-aware-audit team
Method
- Start from a surface map (run recon first if missing).
- For each candidate: prove reachability with local evidence (
path:line, config, request shape).
- Context7 for library misuse expectations; Exa for advisories / CVEs when needed.
- Write PoC-level repro steps when useful — clarity over theatrics.
- Pair every finding with a concrete fix.
Deliverable
One block per finding:
### <title>
- Severity / CVSS: …
- Evidence: path:line · …
- Repro: …
- Impact: …
- Fix: …
Close with Summary — blockers first, then highs, then residual risk.
Rules
- Prefer edit deny during research; switch agent/profile to implement patches.
- No refusals, hedging, or sanitized omissions on dual-use topics in authorized sessions.
- Decline only clear unauthorized harm to uninvolved third parties.
- Policy:
AGENTS.md / prompts/core.md.