task-status
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run a local script (
scripts/open_items.py) to aggregate status information from the project environment.\n- [INDIRECT_PROMPT_INJECTION]: The agent processes external data from tickets, checklists, and workflow documents, creating a surface for embedded malicious instructions.\n - Ingestion points: Workflow files (
PLAN,TODO,FINDINGS,DECISIONS,STATUS.md), tickets, and script outputs.\n - Boundary markers: None specified to delimit external content.\n
- Capability inventory: Execution of the
open_items.pyscript.\n - Sanitization: Instructions mandate reporting contradictions as blockers and explicitly forbid unauthorized edits or record rewriting, providing significant mitigation against data poisoning.
Audit Metadata