delegate-chore

Pass

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest project files and user instructions to generate task requests for subagents (e.g., Claude Haiku). This creates a pipeline where untrusted data from the codebase can influence subagent behavior.
  • Ingestion points: scripts/run.sh and scripts/prepare.sh receive Markdown task definitions containing objectives, scope, and context via stdin.
  • Boundary markers: The skill uses Markdown headers to structure requests, but the provided Bash shims do not perform explicit escaping or sanitization of content passed to the subagent.
  • Capability inventory: The skill can execute Bash and Node.js; subagents invoked by this skill typically possess capabilities for file system manipulation and command execution.
  • Sanitization: Not visible in the shell scripts; sanitization logic is likely deferred to the delegate-cli.mjs tool which is not provided for analysis.
  • [COMMAND_EXECUTION]: The skill's architecture relies entirely on the execution of a Node.js CLI tool (delegate-cli.mjs) through a series of Bash wrappers.
  • Evidence: Every script in the scripts/ directory (e.g., run.sh, dispatch.sh, prepare.sh) uses exec node to invoke the underlying logic. While this is the intended design, it requires the environment to have Node.js 24+ installed and grants the skill the ability to run arbitrary JavaScript logic via the Node.js runtime.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 24, 2026, 01:47 AM