multi-cli-review-action
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external reviewer reports to drive code modifications and execute shell-based verification commands.
- Ingestion points: Processes markdown files found within the specified task directory (
{task-dir}/review-round-{N}/). - Boundary markers: The skill mandates a strict intake contract, requiring specific YAML frontmatter (task-id, round, reviewer-id, etc.) and file naming patterns to identify valid reports.
- Capability inventory: The skill is capable of modifying source code files (applying fixes) and executing shell commands (verification steps).
- Sanitization: The skill requires the agent to re-check all candidate findings against local code and project rules (Step 4) and explicitly mandates user confirmation before any file modifications occur (Step 7).
- [COMMAND_EXECUTION]: The skill involves the execution of external tools for verifying applied fixes.
- Evidence: Step 9 (Verify Before Completion) instructs the agent to run validation commands such as lint, typecheck, or tests.
- Context: While these commands are execution-related, they are part of a standard development workflow and are constrained by the agent's project environment.
Audit Metadata