pattern-creation

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-provided HTML templates to generate pattern files and UX summaries, which creates a surface for indirect prompt injection. Hidden instructions in the source HTML could theoretically attempt to influence the agent's summary generation logic.\n
  • Ingestion points: Phase 1 ("Locate Template") in SKILL.md reads HTML from the conversation context or user input.\n
  • Boundary markers: The skill does not explicitly define delimiters to separate untrusted HTML content from the instruction context during the documentation phase.\n
  • Capability inventory: The skill possesses file-writing capabilities to persist generated patterns to the local filesystem.\n
  • Sanitization: The workflow incorporates a validation step using an internal api_template_validate tool to check HTML structure before persistence.\n- [COMMAND_EXECUTION]: The skill performs file system write operations to persist generated TypeScript code to the local project directory.\n
  • Evidence: Phase 7 ("Write and Verify") describes the writing of .examples.ts files to the projects/internals/patterns/src/ path.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 08:13 AM
Security Audit — agent-trust-hub — pattern-creation