wait-for-ci
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes external CLI tools including
gh(GitHub),glab(GitLab), andvercelto monitor pipeline status and retrieve logs. These tools are expected given the skill's purpose and are executed using parameters derived from the local git environment. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8). It retrieves the last 30 lines of logs from failed CI runs and presents them to the agent for debugging. Since logs can contain arbitrary output from the code being tested, a malicious payload in the codebase could trigger unwanted agent behavior.
- Ingestion points: The
scripts/wait-ci.shscript captures failed logs via thegh run view --log-failedcommand. - Boundary markers: No boundary markers or delimiters are present; the logs are directly interpolated into the JSON
detailsfield. - Capability inventory: The agent has access to the
Bashtool and is explicitly instructed inSKILL.mdto fix code and push changes based on the log analysis. - Sanitization: No sanitization, escaping, or filtering is performed on the captured log content before it is passed to the agent.
Audit Metadata