eda-chief

Legacy compatibility shim for historical eda-chief calls. Use when old prompts/scripts still invoke eda-chief, then dispatch to the repo-level agent policy and execution skills.

EDA Chief

Role

eda-chief is no longer the primary orchestrator. It is a compatibility layer for older workflows that still call this skill name.

Top-level governance is defined in AGENTS.md at repo root.

What to do when invoked

  1. Load AGENTS.md and follow its global routing/governance policy.
  2. Convert legacy "chief-style" request into a scoped execution brief.
  3. Dispatch execution to eda-loop.
  4. Add control skills only when needed:
    • eda-theory-veto for high-cost/high-risk proposals.
    • eda-retro after experiment batches that need next-step decision.
  5. Keep outputs short and artifact-driven.

Hard boundaries

  1. Do not redefine global entry policy inside this skill.
  2. Do not duplicate global recursion/self-update policy from AGENTS.md.
  3. Prefer domain skills for domain reasoning and implementation.
  4. Mention compatibility mode explicitly when this skill is used.

References

Load only when needed:

  1. references/delegation-matrix.md
  2. references/autonomy-guardrails.md