The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes code changes from the repository, which constitutes an ingestion point for untrusted data that could potentially contain instructions intended to influence the sub-agents' analysis.
Ingestion points: Reads code diffs and file contents using git as described in the revision range establishment step.
Boundary markers: Does not include specific prompt delimiters to separate code content from instructions for the sub-agents.
Capability inventory: Uses standard git commands and provides suggestions for manual validation commands.
Sanitization: Content is processed directly for review without specific filtering for embedded prompt instructions.
[COMMAND_EXECUTION]: The skill instructions include standard shell commands for git operations to identify the scope of changes.
Evidence: git merge-base and git diff commands are suggested in SKILL.md for establishing context.

deep-review