The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied proto files and database design documents to generate use case specifications, which introduces a surface where malicious instructions in those files could influence agent behavior.
Ingestion points: Files located in docs/specs/grpc/, docs/specs/data/, and docs/standards/tech-stack-rule.md are read as input for subagents.
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions to isolate content from these external files during processing.
Capability inventory: The skill uses subagents to write and modify files within the docs/specs/grpc-UCS/ and docs/specs/grpc-UCS-review/ directories.
Sanitization: There is no evidence of sanitization or filtering applied to the ingested proto or markdown content before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill involves file system operations to manage documentation and clean up temporary review artifacts.
Evidence: The skill specifically instructs the agent to perform directory creation (mkdir -p) and directory removal (rm -rf equivalent) for managing the docs/specs/grpc-UCS-review/ directory.

ucs-grpc