What the reviewer found
A bug-bounty methodology playbook (recon, IDOR/SSRF/XSS testing, CI/CD attack chains) written for programs the tester is authorized to test, including a rule to verify scope ownership first. The flagged prompt-injection, credential-harvesting and cloud-metadata text describes attack techniques to test FOR (e.g. a firebaseio test write, a documented tj-actions CVE), not something the skill does to the installer; dual-use pentest documentation, truncated past 49KB but consistently descriptive throughout.
dual-use— a security tool that can be misusednetwork— calls the vendor’s APIcredential-access— reads credentials
What was read
The file as it ships in adriannoes/awesome-agentic-ai:
cursor-claude-codex/skills/bug-hunter/skills/bb-local-toolkit/SKILL.md
What the static scan said
The scan flagged 7things. The reviewer kept 0 and dismissed 7 as false.
P1Instruction-override phrasing — false positiveP2Hidden instructions — false positiveE1Sends data to an external URL — false positiveE2Harvests environment variables — false positiveSSRF1Cloud metadata endpoint — false positiveNETMakes network calls — false positiveSHRuns shell commands — false positive
How this review was made
Sonnet 5 read the files above on 7 September 2026 and answered three questions: is it dangerous to whoever installs it, is each scanner finding real, and what should the installer know. The verdict is bound to the file's hash; when the file changes, it is scanned afresh and reviewed again. A script that changes while the definition does not is not re-reviewed — that is a known gap. How the scan and the review work.