bidwriter

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Analysis of external documents presents an indirect prompt injection surface.
  • Ingestion points: The skill ingests untrusted data from tender documents and project specifications (provided via $ARGUMENTS or direct user input) during the 'Project Initialization' phase.
  • Boundary markers: The instructions do not define specific delimiters or 'ignore' directives to isolate the external document content from the agent's instruction set.
  • Capability inventory: The skill facilitates directory creation, file generation (writing the technical and business bid sections), and the invocation of document conversion tools like docx, pdf, and pptx.
  • Sanitization: No evidence of input sanitization or validation is present to detect or strip malicious commands that might be embedded within the processed tender documents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 10:43 AM