The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from plan.md and git diff outputs, which are analyzed by subagents. This creates a surface for indirect prompt injection where malicious instructions embedded in the code or plan could attempt to bias the review results.
Ingestion points: plan.md, git log, git diff scope, and foundation/glossary.md.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands when fanning out to subagents.
Capability inventory: The skill writes findings to impl-review.md, updates metadata in change.md, and executes local git commands.
Sanitization: No evidence of sanitization or escaping is present for the data processed from the plan or the code diff.
[COMMAND_EXECUTION]: The skill uses git log and git diff to retrieve the change history and scope. These are standard development operations required for code review and do not involve shell injection or elevated privileges.

dx-impl-review