The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands for repository analysis, including git diff and git log. User input for the comparison base is resolved via git rev-parse before use.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content into sub-agent prompts, creating a surface for injection.
Ingestion points: Git diffs, commit messages, and external specification files in SKILL.md.
Boundary markers: Not specified; instructions do not explicitly mandate delimiters or instruction-ignoring wrappers for the ingested content.
Capability inventory: Access to local files and shell execution via git commands.
Sanitization: Content from the diffs and specs is interpolated directly into prompts without documented filtering.

code-review