The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted firmware code which presents an indirect prompt injection surface.
Ingestion points: External code is ingested into the agent context via the Read tool and git diff command results as instructed in SKILL.md.
Boundary markers: The instructions lack delimiters (such as XML tags or specific markdown blocks) or warnings to the agent to ignore embedded instructions within the firmware comments or code.
Capability inventory: The skill utilizes Bash(git:*), Read, Grep, and Glob tools to inspect the file system and version history.
Sanitization: No sanitization, escaping, or validation logic is present to handle potentially malicious instructions embedded in the analyzed data.

review-firmware