full-stack-doc
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a data ingestion surface where the agent reads external project files to generate documentation. Ingestion points: The agent is instructed to read source code and manifests from a target repository (SKILL.md). Boundary markers: Explicit instructions in SKILL.md and references/quality-rubric.md require the agent to label assumptions and inferences, reducing the risk of prompt injection from untrusted source data. Capability inventory: The skill utilizes local scripts for structural validation (scripts/validate_templates.py). Sanitization: Not explicitly implemented for the ingestion phase, but the included validation scripts scan outputs for sensitive patterns like hardcoded passwords and private paths.
- [COMMAND_EXECUTION]: The skill requires the execution of local Python scripts (scripts/validate_templates.py and scripts/validate_architecture_filenames.py) to audit the generated documentation. These scripts use standard Python libraries and perform benign validation tasks aligned with the skill's primary purpose.
- [CREDENTIALS_UNSAFE]: While the templates contain placeholders for test environment credentials (e.g., {{TEST_PASSWORD}}), the documentation in references/placeholder-protocol.md provides explicit safety guidelines directing the user to never place real secrets in the templates and to use secret-store references or environment variables instead.
- [SAFE]: The skill provides structured templates and developer-focused guidelines that follow standard industry practices for technical documentation and project management without incorporating malicious code or deceptive instructions.
Audit Metadata