The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to influence the reviewer agents.
Ingestion points: The skill ingests git diffs, issue descriptions, and scope contracts from the repository environment and issue trackers.
Boundary markers: There are no explicit instructions or delimiters provided to the sub-agents to ignore potentially malicious content embedded within the code diffs or issue text.
Capability inventory: The skill uses git, the Task tool for agent dispatch, and executes local shell scripts (reverify-list.sh, review-mark.sh). Note that the skill is restricted via disallowed-tools from using Edit and Write capabilities directly.
Sanitization: No specific sanitization or filtering of the external diff or issue content is mentioned before it is passed to the reviewer agents.
[EXTERNAL_DOWNLOADS]: The documentation suggests the installation of the security-guidance@claude-plugins-official plugin. This refers to an official platform extension provided by a well-known service.
[COMMAND_EXECUTION]: The skill executes maintenance and status-tracking shell scripts located within the plugin's root directory.
Evidence: Executions of ${CLAUDE_PLUGIN_ROOT}/scripts/reverify-list.sh and ${CLAUDE_PLUGIN_ROOT}/scripts/review-mark.sh --set are used to manage the review lifecycle and verify file staleness.

contribution-review