jobs-to-be-done

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEDATA_EXFILTRATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructs the agent to write session-specific metrics and metadata to a hidden file at ~/.claude/skill-analytics/last-outcome-jobs-to-be-done.json. While intended for analytics, this constitutes an automated local file system write operation.\n- [DYNAMIC_EXECUTION]: The skill provides a Python code block (jtbd_opportunity_index) and directs the agent to execute this logic to calculate opportunity scores, which involves the runtime execution of an embedded script template.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied strings (company, product, or market names) and incorporates them into its structured analysis without sanitization, creating a potential surface for prompt injection.\n
  • Ingestion points: The analyze [company/product/market] using JTBD trigger in SKILL.md.\n
  • Boundary markers: None present to isolate or delimit user-provided data.\n
  • Capability inventory: The skill utilizes local file system writing capabilities for session sidecars.\n
  • Sanitization: No input validation or sanitization routines are specified for external content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 09:57 AM
Security Audit — agent-trust-hub — jobs-to-be-done