reversible-agent-task

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests external task briefs and operates on user workspaces, which creates a theoretical surface for indirect prompt injection if the inputs contain malicious instructions.
  • Ingestion points: External data enters the context through the 'task brief' and 'workspace' inputs specified in SKILL_CARD.md.
  • Boundary markers: The skill explicitly mandates a process-level boundary by separating the 'proposal workspace' from the 'target workspace' in SKILL.md, though it does not provide specific data-level delimiters for the task brief content.
  • Capability inventory: The workflow involves execution of implementation loops and verification commands as part of the run and inspect lifecycle steps defined in SKILL.md.
  • Sanitization: The skill focuses on process discipline and does not define specific methods for sanitizing or filtering input data contents.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 08:38 PM
Security Audit — agent-trust-hub — reversible-agent-task