source-driven-development

Pass

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill utilizes the WebFetch tool to retrieve information from highly reputable and well-known official documentation sites, such as react.dev, learn.microsoft.com, and developer.mozilla.org. This practice is documented neutrally as it targets trusted sources to improve implementation accuracy.\n- [PROMPT_INJECTION]: Analysis identified a potential indirect prompt injection surface where external content is ingested into the agent context via documentation fetching.\n
  • Ingestion points: The WebFetch tool in SKILL.md Phase 2 retrieves content from external URLs listed in detect-stack.md.\n
  • Boundary markers: SKILL.md Phase 2/3 classifies fetched content as 'L4 Verify-first' and instructs the agent to perform cross-checks before implementation.\n
  • Capability inventory: SKILL.md Phase 3 empowers the agent to modify project source code based on the fetched data.\n
  • Sanitization: The protocol in SKILL.md Phase 2/3 mandates verification against local code and tests (L3) and explicitly bans high-risk community sources like Stack Overflow or personal blogs to reduce the risk of ingesting malicious payloads disguised as code samples.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 20, 2026, 09:50 PM