cmd-pr-edgecase
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: Employs standard
git diffcommands to retrieve code changes for analysis. This is appropriate for the skill's stated purpose of reviewing Pull Requests. - [DATA_EXPOSURE]: Accesses local project files and version control metadata. No access to sensitive credentials, SSH keys, or environment configuration files was detected.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code content from branch diffs, creating an indirect prompt injection surface.
- Ingestion points: Reads data from
git diffoutput and full file contents via the agent's file system tools. - Boundary markers: None defined to isolate code content from instructions.
- Capability inventory: Analysis is limited to text-based review and reporting. No dangerous capabilities like network exfiltration or shell execution are present in this skill.
- Sanitization: No sanitization or validation of the ingested code content is performed.
Audit Metadata