The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data including code diffs, commit messages, and pull request metadata. This creates an attack surface for instructions embedded within the reviewed data to influence the AI logic.
Ingestion points: External code diffs, commit history, and PR descriptions.
Boundary markers: None; there are no instructions to delimit untrusted input or ignore embedded prompts.
Capability inventory: The skill is instructed to write report files and invoke the issue-writer CLI tool.
Sanitization: No sanitization or validation of the ingested external content is defined.
[COMMAND_EXECUTION]: The skill instructs the agent to execute the issue-writer command-line tool to format the review findings into a formal report.

differential-review