verify-cli-run
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill employs standard system utilities such as git, find, shasum, diff, and grep to verify the state of the local filesystem. It provides a template for wrapping user-initiated commands to capture their output and exit status for later analysis.
- [INDIRECT_PROMPT_INJECTION]: The agent is instructed to ingest and analyze output from external tools and filesystem diffs to reach a verdict. 1. Ingestion points: Output and error streams stored in /tmp/out.txt and /tmp/err.txt, and the results of git status or diff commands, are read back into the agent's context (SKILL.md, steps 4 and 5). 2. Boundary markers: The skill uses a 'pre-defined expectation' strategy (Step 1) to anchor the agent's logic before it sees the tool output, though no technical delimiters are used for the raw tool output. 3. Capability inventory: Verification relies on sh, git, find, shasum, diff, and grep (SKILL.md, steps 2 and 5). 4. Sanitization: External tool output is processed directly by the agent without specific escaping or filtering.
- [EXTERNAL_DOWNLOADS]: The skill suggests using official tools like gh (GitHub CLI) or git ls-remote to verify the state of remote services when verifying network-based operations. These references target well-known, trusted services to confirm the consequence of a command.
Audit Metadata