fable-cowork
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill accepts an
autonomous_taskinput and executes tool chains based on it, creating an ingestion surface for potentially untrusted instructions. However, the skill explicitly defines a multi-stage protocol to lock objectives, enforce safety boundaries, and detect scope drift (Stages 1 and 4) to mitigate these risks. - [COMMAND_EXECUTION]: The skill implements a 'silent execution' policy where tool chains are invoked without mid-chain natural language commentary to reduce developer distractions. While this reduces real-time visibility into individual tool calls, the skill mandates an outcome-first summary with fresh evidence (Stage 8) and provides durable checkpoints to ensure transparency and resumability.
Audit Metadata