ah-create-tasks

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill utilizes standard system utilities like git and basename for environment detection, which is consistent with its role as a development tool.
  • [SAFE]: Workflow state is managed within a vendor-specific directory (~/.agents/arinhub/progresses/), aligning with the author's identity.
  • [SAFE]: The process enforces human-in-the-loop validation during the clarification and complexity check phases, preventing unintended autonomous actions.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project files to drive automation. 1. Ingestion points: prd.md, adr.md, and AGENTS.md. 2. Boundary markers: Absent in subagent instructions. 3. Capability inventory: Command execution (git), file writing, and subagent orchestration. 4. Sanitization: Employs a distillation step to filter technical implementation details from input files before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 01:36 AM