system1-first
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of a local CLI tool (
lp) for intent registry management, evidence logging, and task creation. This is the primary mechanism for its described 'System 1' logic. - [INDIRECT_PROMPT_INJECTION]: The skill defines a process for transforming untrusted user requests into shell command arguments.
- Ingestion points: User-provided business requests and queries are processed in Step 1.
- Boundary markers: The skill explicitly instructs the agent to 'normalize' the intent into a short, stable string without parameter values before use, which acts as a sanitization boundary.
- Capability inventory: The skill uses
lpCLI tools for searching registries, adding evidence, and creating tasks. - Sanitization: The normalization step (e.g., converting 'How much did revenue drop?' to 'weekly revenue by region') prevents direct interpolation of raw user text into command lines.
Audit Metadata