workthrough

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFECREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill encourages documenting configuration changes and environment variables as seen in EXAMPLES.md and TEMPLATE.md. This presents a risk of accidental secret exposure if an agent captures real credentials from a .env file or configuration update and saves them to the repository's documentation directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from code modifications to generate summaries.
  • Ingestion points: All files modified during the development session (referenced in SKILL.md Automation Instructions).
  • Boundary markers: None identified; the instructions do not provide delimiters to isolate analyzed code from the documentation prompt.
  • Capability inventory: File writing to the local workthrough/ directory.
  • Sanitization: None identified; external content from code changes is interpolated into documentation without escaping or filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 12:52 PM
Security Audit — agent-trust-hub — workthrough