The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's relay script (scripts/relay.mjs) executes external CLI tools including cline and git to perform coding tasks. It implements safety measures by validating shell-sensitive arguments (--model, --provider) against a restricted character set and passing the task brief via stdin rather than command-line arguments to prevent shell injection on Windows systems where the shell shim is used.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for an external coding agent, ingesting untrusted task instructions (the "brief") and passing them to an agent with system-level shell and file access. This ingestion point at brief.txt or stdin creates a surface for indirect injection. To mitigate this risk, the skill includes a static boundary prompt ("Follow the task instructions provided on stdin.") and extensive documentation on "Review and Land" procedures, mandating that the orchestrator verify all code changes manually and re-run project gates rather than trusting the agent's self-reported success.
[DYNAMIC_EXECUTION]: The relay script dynamically locates and executes a sibling script (../../delegate-setup/scripts/lane.mjs) using node when the --lane configuration option is active. This execution uses a path relative to the skill's installation directory to resolve configuration dials and represents a local functional dependency within the skill suite.

cline-delegate