The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from a user's codebase (Typescript files) to generate configuration. This introduces an indirect prompt injection surface where malicious instructions could be embedded within code identifiers or comments.
Ingestion points: Uses Read and Bash (via grep) to scan files in the src/ directory.
Boundary markers: Does not use technical delimiters or markers for the interpolated code, but relies on natural language instructions to define the scope of analysis.
Capability inventory: The agent has permission to Write and Edit files and execute rudder-cli commands.
Sanitization: The skill contains a 'Handling External Content' section that explicitly instructs the agent to only parse for type definitions, never execute or evaluate code, and ignore suspicious or obfuscated patterns.

rudder-code-first-instrumentation