The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git and gh (GitHub CLI) commands to resolve target diffs, fetch pull request details, and post comments. These operations are performed in response to user-initiated review requests and are necessary for the skill's stated purpose of reviewing code changes.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data in the form of code diffs and repository files. This attack surface is mitigated by strict instructional boundaries in the agent briefs (found in references/lenses.md and references/workflow-template.md) which explicitly state that the code and diffs are data, not instructions, and forbid the agents from modifying files.
[SAFE]: The skill incorporates safety mechanisms such as adversarial verification (where an independent node attempts to refute findings), scoped contexts for sub-agents (lenses), and requiring explicit user authorization via a token before performing side-effecting operations like posting to GitHub.

panel-review