test
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious behaviors, obfuscation, or exfiltration techniques were found. The skill configuration properly restricts the shell environment to a specific set of tools.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external repository contents and test artifacts, which constitutes a vulnerability surface. 1. Ingestion points: Reads
package.json, test failure outputs, and runner traces such as Playwright screenshots (SKILL.md). 2. Boundary markers: Includes a specific instruction to treat test output as data and not as instructions (SKILL.md). 3. Capability inventory: Permits bash execution forgit,bun, andbunxand allows file edits for authorized repairs (SKILL.md). 4. Sanitization: Implements an authorization gate requiring explicit user permission before any modifications to the codebase are made based on the ingested data.
Audit Metadata