work-review
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill facilitates the execution of repository-specific commands to verify code integrity.
- Evidence: The REFERENCE.md file defines a protocol for running a verify command retrieved from the repository's configuration or manifest files (e.g., package.json).
- Context: This capability is central to the skill's purpose. To prevent abuse, the instructions provide a comprehensive 'Separate verification from operator authority' framework, directing the agent to use sandboxed environments with restricted network and filesystem access.
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from issue trackers and pull requests.
- Ingestion points: Issue bodies and pull request comments across GitHub, GitLab, and Linear as described in SKILL.md and REFERENCE.md.
- Boundary markers: The skill uses an explicit distinction between information used for context and information used for instruction.
- Capability inventory: Execution of shell commands via Bash and forge operations via gh and glab CLIs.
- Sanitization: The 'Author Authority' section in REFERENCE.md establishes a robust permission-checking mechanism. The agent is instructed to verify that the author of any directive has the necessary repository permissions (e.g., 'write' or 'maintainer' roles) before acting on it, effectively neutralizing unauthorized injection attempts.
Audit Metadata