babysit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple shell commands to manage its environment. It uses
nodeto parse version metadata fromversions.json,findto check for leftover work directories in.a5c/runs, andnpmto install or execute the@a5c-ai/babysitter-sdk. It also executes thebabysitterCLI to fetch runtime instructions. - [EXTERNAL_DOWNLOADS]: The skill installs the
@a5c-ai/babysitter-sdkand@a5c-ai/babysitterpackages from the npm registry if they are not already present or if the global shim is broken. These are vendor-owned resources matching the skill author's identity. - [INDIRECT_PROMPT_INJECTION]: The skill retrieves its primary operational instructions by executing
$CLI instructions:babysit-skill. This creates an ingestion point where the output of an external command (the vendor CLI) provides the instructions the agent must follow for orchestration. This is a design pattern for dynamic orchestration rather than a malicious injection.
Audit Metadata