The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external sources, including user-defined goals and automated critique sources (test results, subagent reviews, or evaluation framework scores). This creates a potential surface for indirect prompt injection if the ingested data contains malicious instructions designed to hijack the agent's control flow.
Ingestion points: The workflow ingests external data at the Goal definition and the Critique source evaluation steps in SKILL.md.
Boundary markers: The instructions do not mandate the use of delimiters or specific "ignore embedded instructions" warnings when handling output from tests or subagents.
Capability inventory: The skill possesses capabilities to write iteration artifacts to the .coord/ directory, update a shared .coord/memory.yml file, and spawn subagents (e.g., code-reviewer).
Sanitization: There are no explicit instructions for sanitizing, escaping, or validating the content of external critique results before they influence the next iteration's plan.
[COMMAND_EXECUTION]: The workflow involves the orchestration and execution of delegate tools (Claude, Codex, Gemini) and the spawning of subagents to perform code reviews and tests, which represents an instruction set for multi-step command and tool execution.

agent-plan-act-reflect