full-stack-doc
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: LOWNO_CODE
Full Analysis
- [PROMPT_INJECTION] (SAFE): No patterns indicating attempts to override agent instructions or bypass safety filters were found in the skill documentation or metadata.
- [DATA_EXFILTRATION] (SAFE): No network-related commands (curl, wget) or access patterns to sensitive files (e.g., ~/.ssh) are present.
- [REMOTE_CODE_EXECUTION] (SAFE): The skill is strictly informational and does not contain dependencies or commands for downloading and executing remote code.
- [INDIRECT_PROMPT_INJECTION] (INFO): The skill defines an attack surface by populating templates with user-provided data. 1. Ingestion points: Placeholders such as {产品名称} and {项目名称} in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Text generation and display only. 4. Sanitization: Absent. The severity is INFO because the capability is limited to informational text generation.
Audit Metadata