ship
Warn
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the 'gh pr merge --admin' command, which allows the agent to bypass standard repository branch protections such as mandatory reviews or status checks. This could potentially allow unverified or unauthorized code to be merged into protected branches.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is instructed to process untrusted data from pull request comments and review feedback. Since the agent also has repository write access and the capability to execute local scripts, an attacker could use PR comments to steer the agent into performing unauthorized actions.
- [COMMAND_EXECUTION]: The skill executes the 'pnpm run prep' command to validate changes, which involves running arbitrary scripts defined in the project's local configuration.
Audit Metadata