The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes git commands (git diff, git log, git rev-parse) to compare branches and retrieve commit history. This usage is transparent and directly related to the tool's purpose of code review.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to the processing of data from external sources and repository contents.
Ingestion points: The skill ingests untrusted data from git diff output, git log commit messages, and external content from issue trackers (Linear, ClickUp) or specification files (e.g., in docs/ or specs/).
Boundary markers: The sub-agent prompts lack explicit delimiters or instructions to treat the included diffs and spec data as non-executable text, which could allow instructions within those data sources to influence the sub-agent's behavior.
Capability inventory: The skill has the ability to execute shell commands (git) and spawn sub-agents using the Agent tool.
Sanitization: No validation or sanitization of the external data or diff content is specified before interpolation into the sub-agent prompts.

code-review