implement
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious code, hardcoded credentials, or suspicious network activities were detected. The skill's instructions are consistent with its stated purpose of implementing code changes.- [PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection due to its core functionality of processing external code. An attacker could embed instructions in the codebase to manipulate the agent during the implementation or verification steps.
- Ingestion points: Reads project files during the code inspection phase (SKILL.md).
- Boundary markers: None identified; there are no instructions to ignore embedded commands in the files being processed.
- Capability inventory: The skill explicitly authorizes the agent to modify the file system ("change code") and execute shell commands ("Run relevant verification") (SKILL.md).
- Sanitization: There are no documented steps for validating or sanitizing the content of the project files before processing or execution.
Audit Metadata