warisskill-research-technical

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a research methodology that involves ingesting data from external, untrusted sources such as GitHub Issues, Discussions, and community blog posts. This creates a potential surface where malicious instructions hidden in external technical content could influence the agent's research findings or subsequent implementation steps.
  • Ingestion points: Technical documentation, source code repositories, GitHub Issues, Discussions, and community websites (Stack Overflow, blogs).
  • Boundary markers: Absent. While the skill establishes a trust hierarchy to prioritize sources, it does not instruct the agent to use specific technical delimiters or boundary markers when interpolating this external content into its context.
  • Capability inventory: The skill is intended for use during engineering "spikes" and technical evaluation phases, where the agent typically possesses capabilities for file modification and command execution within the development environment.
  • Sanitization: No explicit instructions are provided for filtering, escaping, or sanitizing content retrieved from community sources before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:16 PM
Security Audit — agent-trust-hub — warisskill-research-technical