request-plan-review
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
command -vto detect the presence of local command-line tools likeclaude,gemini,codex,opencode, andqwen. It subsequently dispatches subagents to execute these detected CLIs to perform automated reviews of the source markdown. - [EXTERNAL_DOWNLOADS]: The generated HTML review page is configured to fetch client-side libraries for syntax highlighting, diagram rendering, and visual annotations from well-known and trusted CDNs, including ESM.sh, JSDelivr, and Unpkg.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through its markdown processing workflow.
- Ingestion points: The skill reads user-provided or agent-generated markdown plan files from the local filesystem.
- Boundary markers: The
buildReviewPromptutility inscripts/review-utils.mjsinterpolates the full markdown content into a review prompt without utilizing defensive delimiters or specific 'ignore' instructions for the reviewer subagents. - Capability inventory: The tool possesses capabilities for file system writes (
docs/request-plan-review/), command execution (command -v,open), and the ability to trigger external CLI tools via subagents. - Sanitization: There is no evidence of filtering or escaping performed on the markdown content before it is processed by the reviewer agents.
Audit Metadata