planning-and-task-breakdown

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external specifications and codebase content to generate implementation plans and task lists.
  • Ingestion points: The planning process begins by reading the spec and codebase sections (SKILL.md, Step 1).
  • Boundary markers: There are no explicit delimiters defined to separate agent instructions from the content of the external specifications.
  • Capability inventory: The skill writes to the filesystem (tasks/ directory) and can interact with external issue trackers using CLI tools (e.g., bd).
  • Sanitization: There is no validation or sanitization of input specifications before they are used to populate task descriptions, acceptance criteria, or verification commands.
  • [SAFE]: The skill implements a safety check to prevent overwriting existing plan or todo files if they contain incomplete tasks, requiring user confirmation for such actions.
  • [SAFE]: The skill refers to well-known external services (GitHub Issues, Jira, Linear) for task tracking, which is standard professional practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 07:21 PM
Security Audit — agent-trust-hub — planning-and-task-breakdown