aep
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes standard developer tools including
gitandghfor version control and issue management. These tools are pre-authenticated by the platform, and the skill explicitly forbids the agent from attempting to manage credentials or authentication manually. - [DATA_EXFILTRATION]: The skill establishes strict filesystem boundaries, prohibiting access to files outside the project directory (e.g.,
~or other projects). It also includes a rule against including secrets in search queries or URLs. - [PROMPT_INJECTION]: The skill contains explicit instructions to treat web results and fetched pages as untrusted data rather than instructions, providing a defense against indirect prompt injection from external sites.
- [REMOTE_CODE_EXECUTION]: While the skill allows the use of project-specific package managers (npm, go, bal), it restricts the installation of global tools (brew, apt, npm -g) and limits execution to compilation/verification steps, explicitly forbidding the execution of the application itself in the sandbox.
- [EXTERNAL_DOWNLOADS]: The skill interacts with GitHub for repository operations and issue fetching. GitHub is a well-known service, and the operations are performed using authenticated platform tools, which is consistent with the skill's purpose as a coding assistant.
Audit Metadata