Context Management for Agents — Build Runbook

End-to-end plan to take the project from a skills collection to an enterprise platform, in four phases.

Current state

Phase	Status
0 — Skills + validation OS	Done
1 — Consumable runtime	Done
2 — Operationalized primitives	Done
3 — Enterprise hardening	Done
4 — Scale & ecosystem	Planned

See Roadmap for Phase 4 scope.

Standing guardrails

1. Never break the four validation gates before commits.
2. Do not fork router/primitive logic — one implementation in runtime/core/.
3. Do not change skill content to make code pass.
4. Honest benchmarks, estimated costs, no certification claims.
5. Preserve MIT attribution chain.
6. Deterministic offline tests (mock IdP/OPA/vault).
7. Checkpoint commits; no push unless explicitly requested.
8. Each phase lists out-of-scope items as hooks only.

Four gates

python researcher/scripts/validate_repo.py --strict
python researcher/scripts/skill_health.py --strict --no-history
python researcher/scripts/run_benchmarks.py
python researcher/scripts/check_activation_cases.py

Runnable proof

python examples/demo/run_demo.py
python examples/demo/run_mcp_demo.py