antithesis-skills-feedback
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute
git rev-parse --short HEADwithin the local skills repository. This is used solely to identify the current version/commit SHA for accurate bug reporting. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for user feedback and agent self-assessments, which are interpolated into a GitHub issue URL.
- Ingestion points: User-provided summaries of issues (
what-happened) and agent-generated self-assessments (agent-explanation). - Boundary markers: The skill uses URL query parameters as structural boundaries for the ingested data.
- Capability inventory: The skill has the capability to execute shell commands (
git) and display URLs to the user. - Sanitization: The instructions explicitly require the agent to percent-encode all values before URL construction and mandate that the final URL must be presented for user review rather than auto-submitted, serving as a critical human-in-the-loop safety control.
Audit Metadata