Four Collaboration Zones

david-sanchez's model for structured human-agent collaboration across the SDLC. Unstructured collaboration "leads to inconsistent outputs, duplicated work, and trust erosion." Structured collaboration is about designing the interaction model so both contribute their respective strengths within a shared framework of accountability.

The zones

1. IDE / Editor — human defines intent; agent drafts completions/tests. Governance: real-time accept/reject + editor-level context files.
2. Pull Request — human reviews and approves; agent opens PRs and responds to review comments. Governance: branch protection, required approval, agent-specific labels.
3. CI/CD Pipeline — human defines rules and reviews failures; agent triggers builds and remediates in scope. Governance: agent verification layers, scope validation, provenance checks.
4. Production — human monitors and owns incident response; agent detects anomalies and executes pre-approved remediations. Governance: runbook automation + human approval gates for high-risk actions.

The key insight

"Agents operate best when they have clearly defined scope, structured inputs, and explicit governance boundaries. The collaboration is not about letting agents loose."

Connects to

• delegated-agent-authorization — Partee's auth layer is exactly how you enforce scope at zones 3–4.
• coding-agent-as-foundation — Chase's thesis extends zones 1–2 to general-purpose work.
• pipeline-as-verifier — zone 3 mechanics.