Skills
Navigate
skills/rheinmir/skills/failure-flywheel/Gen Agent Trust Hub

failure-flywheel

Pass

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted failure summaries and details which are persisted and later re-injected into the agent's context during the drafting and proposal workflow.
  • Ingestion points: The record command in SKILL.md accepts <summary> and --detail arguments directly from the agent's failure context.
  • Boundary markers: The skill lacks explicit sanitization or delimiters to isolate logged failure descriptions from the rest of the draft proposal content.
  • Capability inventory: The skill performs file-write operations to failures.jsonl and the llmwiki/wiki/sources/draft/ directory.
  • Sanitization: No escaping or validation of the failure summary content is mentioned before it is interpolated into the markdown draft stubs.
  • [COMMAND_EXECUTION]: The skill relies on the execution of local Python scripts to process failures and manage the taxonomy.
  • Evidence: SKILL.md documents the use of python3 harness/scripts/failure-flywheel.py and python3 harness/scripts/flywheel.py for recording, reporting, and drafting operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 14, 2026, 02:24 AM