What the reviewer found
This file lives at examples/malicious-example/SKILL.md and its own banner says it is a TEST FIXTURE for the skills-guard scanner, warning not to install or invoke it against a real agent. It intentionally packs a prompt-injection payload, SSH-key exfiltration, a zero-width-space-obfuscated rm -rf, and fake credentials so the scanner has something real to catch — a demonstration corpus, not a live attack shipped for installation.
prompt-injection— tries to steer the agentexfiltration— sends data outcredential-access— reads credentialsobfuscated— hides what it runsdual-use— a security tool that can be misusedexample-or-fixture— an example or test fixture
What was read
The file as it ships in aibuild-lab/skills-guard:
examples/malicious-example/SKILL.md
What the static scan said
The scan flagged 7things. The reviewer kept 7 and dismissed 0 as false.
P1Instruction-override phrasing — realP2Hidden instructions — realAR3Nullifies safety policies — realE1Sends data to an external URL — realPE2Asks for root — realPE3Reaches for credential files — realNETMakes network calls — real
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.