The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions involve analyzing existing middleware and logging setups using Grep and Read to identify observability gaps. This introduces an attack surface where malicious comments or instructions embedded in the analyzed codebase could potentially influence the agent's behavior.
Ingestion points: SKILL.md instructions specify using Grep and Read on the target application's source code.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for these read operations.
Capability inventory: The skill has access to Write, Edit, and a restricted Bash command for project monitoring tasks.
Sanitization: There is no explicit validation or sanitization of the content read from existing project files.
Note: This is a common surface for development-focused skills, and the instructions provide strong context that minimizes the risk of the agent obeying external instructions.

monitoring-apis