The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files and git diffs, creating a surface for indirect prompt injection where malicious instructions embedded in code could influence agent actions.
Ingestion points: The skill reads file content and git diff outputs in SKILL.md Step 1 and Step 3.
Boundary markers: The skill does not implement specific boundary markers or 'ignore embedded instructions' warnings for the data it processes.
Capability inventory: The skill possesses the capability to write to the local file system (Step 5) and execute arbitrary shell commands for running test suites (Steps 2 and 6).
Sanitization: There is no evidence of data sanitization or validation of the content ingested from the repository.
[COMMAND_EXECUTION]: The skill identifies and executes test suites based on repository content.
Evidence: Steps 2 and 6 in SKILL.md instruct the agent to 'Find and run the suite covering the changed files' and record the command for later verification. This involves executing potentially untrusted shell commands discovered within the project environment.

polish-diff