use-wayfinder

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the knowledge/ directory within a project to provide grounded answers.
  • Ingestion points: The agent reads and searches through files located in the knowledge/ bundle (or path specified by WAYFINDER_KNOWLEDGE_DIR).
  • Boundary markers: Absent. There are no explicit instructions or delimiters provided to the agent to disregard potential instructions embedded within the knowledge files being searched.
  • Capability inventory: The skill executes the wayfinder CLI for validation, indexing, and searching. It also modifies local configuration files (e.g., .mcp.json) via wayfinder setup and writes derived index data to the machine.
  • Sanitization: None detected. The skill relies on the semantic search results of the Wayfinder tool without explicitly sanitizing or filtering the content for prompt injection patterns before inclusion in the LLM's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 09:11 PM
Security Audit — agent-trust-hub — use-wayfinder