architect-python-uv-takehome

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the creation of document extraction pipelines that ingest untrusted data, representing a surface for indirect prompt injection. \n
  • Ingestion points: Document processing logic located in the src/extractor/ directory, as specified in SKILL.md. \n
  • Boundary markers: The scaffold does not include predefined delimiters or safety instructions to prevent the agent from obeying commands embedded within the documents it parses. \n
  • Capability inventory: Generated projects have the capability to write to the local filesystem (into the output/ directory) and can optionally interact with LLM providers using the openai library. \n
  • Sanitization: The skill provides no default mechanisms for sanitizing or validating extracted text before it is utilized in downstream agent logic or narrative reporting.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 12:40 PM