maintenance-pr
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external diagnostic findings from files in the
.context/directory. While these files are intended to be generated by other trusted agents, they represent a potential attack surface if the source data is compromised. - Ingestion points: Reads diagnostic YAML files from
.context/via thefinding_fileinput. - Boundary markers: Lacks explicit delimiters for untrusted data, but enforces a strict allow-list of change types and recognizes specific actions.
- Capability inventory: Includes file system modification, Git commits, pushing to remote repositories, and creating GitHub pull requests via
gh pr create. - Sanitization: Aborts operations if "credential-shaped strings" are detected in the generated diffs.
- [COMMAND_EXECUTION]: The skill utilizes several CLI tools to perform its tasks, specifically
git,gh(GitHub CLI), and thebruinCLI. These are restricted to repository management and interactions with legitimate services associated with the skill's purpose.
Audit Metadata