The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository and external issue trackers which could contain embedded instructions.
Ingestion points: Untrusted data is collected via git diff, git log, and from files in docs/, specs/, or .issues/ directories, as well as fetched issue tracker data.
Boundary markers: The process uses sub-agents to separate the review axes, but the instructions lack specific prompt delimiters (like XML tags or triple quotes) or 'ignore' instructions for the external content passed to these agents.
Capability inventory: The skill uses the git CLI and sub-agent orchestration.
Sanitization: No explicit sanitization or validation of the ingested diff content or spec text is mentioned before interpolation into prompts.
[COMMAND_EXECUTION]: The skill utilizes git diff, git log, and git rev-parse to perform its primary function of code analysis. These commands are used within the scope of a development repository.

code-review