architecture
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
git config user.nameto identify the author of the document. This is a standard development practice for automated documentation and does not involve unsanitized user input or risky shell commands. - [INDIRECT_PROMPT_INJECTION]: The skill ingests content from project files like
REQUIREMENTS.mdand source code. While these files are external to the skill itself, the skill's environment is restricted viadisable-model-invocation: true, which prevents the agent from being manipulated into calling other tools or skills if malicious instructions are found in the ingested data. - [SAFE]: The skill implements a clear human-in-the-loop requirement, explicitly stating that the document status should only be set to 'Approved' when a human approves it. It lacks network capabilities and does not access sensitive system paths or credentials.
Audit Metadata