The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using node and the agy CLI to delegate coding tasks to a separate implementer agent.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating a potential surface for indirect prompt injection attacks.\n
Ingestion points: External brief.txt files (task instructions) and the finalMessage property of result.json which contains the output from the delegated agent.\n
Boundary markers: The documentation includes detailed manual review checklists in references/review-and-land.md to verify all implementer output before committing.\n
Capability inventory: The skill has the capability to execute shell commands (node, agy) and modify the local filesystem via git.\n
Sanitization: There is no automated sanitization of incoming briefs; the skill relies entirely on human-in-the-loop review and project-specific testing gates.\n- [EXTERNAL_DOWNLOADS]: The skill documentation describes the use of a helper script, scripts/relay.mjs, which is not bundled with the provided files and must be downloaded from a third-party community repository (amElnagdy/delegate-skills).\n- [COMMAND_EXECUTION]: The skill documentation includes the --dangerously-skip-permissions flag, which is designed to bypass security and permission prompts within the implementer's execution environment.

agy-delegate