craft-autoresearch
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute a 'run harness' command provided by the user and several git commands (commit, reset, restore) to facilitate the optimization loop. This is a functional requirement of the tool.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it requires the agent to process and score untrusted output from external artifacts. * Ingestion points: The agent analyzes output from the 'run harness' command (SKILL.md, Steps 3 and 4). * Boundary markers: There are no instructions to use markers or to ignore instructions embedded in the harness output. * Capability inventory: The agent has permissions to execute shell commands, perform git operations, and write to the ~/.craftkit/ directory. * Sanitization: The skill does not describe any validation or sanitization of the artifact output before it is evaluated by the agent.
Audit Metadata