wk-workstyle-error-handling

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes project files which could potentially contain indirect prompt injection attacks designed to mislead the agent.
  • Ingestion points: File content is ingested via Read, Glob, and Grep tools as defined in SKILL.md.
  • Boundary markers: There are no explicit instructions or delimiters used to separate untrusted code data from the agent's core instructions.
  • Capability inventory: The skill is restricted to read-only tools and does not have the ability to write to the filesystem or access the network.
  • Sanitization: The skill does not perform any escaping or validation of the content read from the source files.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 09:50 PM
Security Audit — agent-trust-hub — wk-workstyle-error-handling