The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by reading and processing local file contents which could be controlled by an attacker.
Ingestion points: The skill identifies and reads modified API definition files (e.g., .proto, .yaml, .go) using the git diff command as described in SKILL.md.
Boundary markers: No specific boundary markers or 'ignore' instructions are used in SKILL.md to isolate the contents of the files being analyzed from the agent's control flow.
Capability inventory: The skill employs the Bash tool to execute git, buf, and oasdiff commands (SKILL.md), providing a mechanism where instructions in the data could potentially influence shell operations.
Sanitization: There is no evidence of sanitization or filtering of the file contents before they are analyzed and reported by the agent.

api-compat