plan-tasks

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill follows its stated purpose of assisting in task decomposition. It reads local files and writes task documentation without performing network requests or sensitive data exfiltration.
  • [PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection as it processes untrusted content from documentation files to generate subagent tasks.
  • Ingestion points: Processes proposal.md, design.md, and specs/ files from the change directory (SKILL.md, Step 1).
  • Boundary markers: Does not utilize specific boundary markers or 'ignore' instructions to isolate input data from the prompt logic when synthesizing the task files.
  • Capability inventory: Capable of creating and writing multiple markdown files within the tasks/ subdirectory and the tasks.md file (SKILL.md, Step 4 and 6).
  • Sanitization: Content from specification scenarios is copied verbatim into generated task files without sanitization or escaping (SKILL.md, Step 4).
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 04:57 PM
Security Audit — agent-trust-hub — plan-tasks