publish-review
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes findings from code review reports, which could theoretically contain malicious instructions if the source code or review data is adversarial.
- Ingestion points: The skill ingests data from a previously generated
/code-reviewreport in the conversation (SKILL.md, Step 1). - Boundary markers: The skill enforces a mandatory human approval requirement in Step 3, where the user must explicitly confirm the final draft before any network or CLI operations occur.
- Capability inventory: The skill has the capability to publish data to external trackers using MCP tools or system CLI commands (SKILL.md, Step 4).
- Sanitization: It utilizes an optional
unslopskill to clean and refine the drafted text (SKILL.md, Step 2). - [COMMAND_EXECUTION]: The skill instructions mention using a 'verified CLI' as a fallback mechanism for publishing reviews if MCP tools are unavailable.
- Evidence: Step 4 of the process describes using a verified CLI or platform UI to submit review bodies and inline suggestions.
Audit Metadata