agentleFS
aaronreneaaronrene/knowtation/.cursor/rules/orchestrator.mdc

knowtation / rules

Cursor rule1 starsChanged 2 months agoMITView on GitHub

RULE #8 — Orchestrator via Overseer Kit (roadmap + handover day-to-day)

The file

We split this orchestrator.mdc into the rules it gives an agent

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When this repo has .overseer/config.yaml, run phased work through the kit — do not invent a parallel hand-rolled ROADMAP / HANDOVER …

    line 9
  2. Very specific:

    Every roadmap row and handover NEXT / paste block must include Model: with exactly one of:

    line 46
  3. Distinctive:

    Never Outline/Plan on Auto.

    line 55
  4. Very specific:

    pass: Never mark DONE on green tests alone.

    line 56
  5. Very specific:

    Read {{docs.roadmap_path}} (target phase) and {{docs.handover_path}} (NEXT + verified snapshot)

    line 61
  6. Very specific:

    Prefer kit rules/skills (.cursor/rules/*, .cursor/skills/*, .claude/skills/*) over folklore

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 6,668 Cursor rule files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
orchestrator.mdc
---
description: RULE #8 — Orchestrator via Overseer Kit (roadmap + handover day-to-day)
alwaysApply: true
---

# RULE #8 — Orchestrator (Overseer Kit)

**Source of truth:** [🆗 Overseer Kit](https://github.com/aaronrene/overseer-kit).
When this repo has `.overseer/config.yaml`, run phased work **through the kit** — do not invent a
parallel hand-rolled ROADMAP / HANDOVER protocol.

## Install and refresh

| Situation | Command |
| --- | --- |
| Greenfield (no kit yet) | `ok init --regime <git-only\|muse+git-mirror\|muse-only>` (or `init --migrate` if living docs already exist) |
| Pull kit updates | `ok sync` |
| Health / drift | `ok status` (use `--exit-code` when gates matter) |
| Docs vs VCS hygiene | `ok governance-sync --dry-run` then apply when correct |
| Freeze gate | `ok review --freeze <artifact>` |
| Ad-hoc honesty | `ok check-ok` / `/check-ok` |

Resolve the `ok` shim from the kit checkout (`./cli/ok` or a path to it). Prefer `ok` over the
deprecated `overseer` compat shim.

Open the repository root (the folder that contains `.overseer/`) as the IDE workspace. If you open
a parent folder, project rules and skills under `.cursor/` often do not load. The CLI still works.
The open editor tab is not the source of truth — run `ok next`.

**Doc paths are config-driven** — read `docs.roadmap` / `docs.handover` from `.overseer/config.yaml`
(here: `{{docs.roadmap_path}}`, `{{docs.handover_path}}`). Do not assume root `ROADMAP.md` /
`OVERSEER_HANDOVER.md` unless config says so.

## Two durable docs (always together)

| Doc | Role |
| --- | --- |
| `{{docs.roadmap_path}}` | Phase control: Phase Model Key, build queue (phase → **Model** → Status → deliverable), Definition of Done |
| `{{docs.handover_path}}` | Session relay: **NEXT SESSION**, paste-ready prompt, verified snapshot, VCS table, change log (KH1 shape) |

Trust the handover **verified snapshot** over chat history. Update both docs in the same closing
commit (SD-17). `/governance-sync` automates the hygiene check — it does not invent policy.

## Model labels (`policy/model-labels.yaml`)

Every roadmap row and handover NEXT / paste block **must** include **`Model:`** with exactly one of:

| Label | Meaning |
| --- | --- |
| **Thinking** | Design + freeze spec (interfaces, fail-closed rules, test matrix) — no mechanical build |
| **Auto** | Build exactly against a frozen spec — no redesign |
| **Thinking → Auto** | Split into `{step}a` (Thinking) then `{step}b` (Auto) — never one combined prompt |
| **Operator + Auto** | Human for Tier-3 live/merge gates; Auto for implementation + doc sync |

Never Outline/Plan on **Auto**. Never start **Auto** until the consumed freeze has review verdict
**`pass`**. Never mark **DONE** on green tests alone.

## Starting a phase

1. Fresh chat with the **Model** tier named in the handover
2. Read `{{docs.roadmap_path}}` (target phase) and `{{docs.handover_path}}` (NEXT + verified snapshot)
3. Paste the **Paste-ready prompt** fence (self-contained; no prior chat required)
4. **Thinking:** freeze WHAT/HOW in a `frozen: true` artifact; run `/freeze-review-loop` → `pass`
5. **Auto:** implement mechanically; run seven-tier tests (`policy/test-tiers.yaml`); run
   `/build-verification-review` → `pass` before ROADMAP → **DONE**
6. Close: update roadmap status + regenerate handover NEXT together; feature-branch commit; PR per
   `{{vcs.regime}}` — **no** merge to `{{vcs.git.main_branch}}` without Tier 3. Final reply
   includes CURRENT NEXT from disk (`ok next` / `## CURRENT NEXT — paste this`).

## Definition of Done (every phase)

- Deliverables match the frozen spec
- Freeze review **`pass`** before Auto that consumes the freeze
- Required seven-tier tests green locally
- Build verification **`pass`** after Auto before status → **DONE**
- No secrets committed
- Both governance docs updated together
- Feature-branch hygiene complete; merge remains Tier 3

## Agent behavior

- Prefer kit rules/skills (`.cursor/rules/*`, `.cursor/skills/*`, `.claude/skills/*`) over folklore
- One **THE ONE NEXT STEP** at a time in the handover
- Dirty tree at session end is a failure — commit on the feature branch (Tier 1)
- Hard stops: secrets, live posture flips, real money, merge to `{{vcs.git.main_branch}}` without operator authorization

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.