Ultra orchestrator
Skill keithhegit/ultra-orchestration/skills/ultra-orchestrator
Primary Ultra Orchestration entry point for strict Codex-controlled engineering runs. Use this single skill for development, bug fixes, OpenSpec work, multi-file implementation, review/QA-heavy tasks, risk-gated work, or any run that needs automatic mode selection, OpenSpec bootstrap, machine-checkable TaskManifest and WorkPackage artifacts, ledger tracking, slice DAG execution, review, QA, delivery evidence, and retro.From its SKILL.md
npx -y skills add keithhegit/ultra-orchestration --skill ultra-orchestratorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
- runs commandsInstructs the agent to run 2 commands, including `scripts/new_run.py` and 1 more.
SKILL.md
6.7 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it
Ultra Orchestrator
Use this skill as the default public entry point for Ultra Orchestration.
The user should only need to invoke $ultra-orchestrator <task description>.
This is the mainline successor to the vNext protocol. The old stable workflow
is preserved as ultra-orchestrator-legacy.
Startup Contract
Recommended invocation:
$ultra-orchestrator <task description>
OpenSpec invocation:
$ultra-orchestrator OpenSpec change <change-id or path>: <task description>
Compatibility invocation:
$ultra-vnext-core <task description>
If the client exposes slash aliases, /ultra-orchestrator can be used the same
way.
Run Mode Decision
Classify the run before planning or execution:
LIGHTreview-only, QA-only, explanation-only, or user-explicit lightweight workSTANDARDuser-explicit fast orchestration with bounded risk and no full control planeSTRICTdevelopment work where OpenSpec is unavailable but ledger and JSON contracts are still requiredSTRICT_OPENSPECdefault for development work, bug fixes, multi-file implementation, pilots, full orchestration tests, OpenSpec work, slice DAG work, and control-plane validation
Development tasks must prefer STRICT_OPENSPEC. If that mode is impossible,
explain why and downgrade explicitly. Markdown-only artifacts are not enough
for STRICT or STRICT_OPENSPEC.
Router Duties
When invoked:
- classify the task and choose
run_mode - create or require OpenSpec change scaffolding when
STRICT_OPENSPECapplies - initialize the run ledger for
STRICTandSTRICT_OPENSPEC - route through the minimal required sibling skills
- validate machine-checkable artifacts before delivery
- produce delivery artifacts or a clear blocker
Do not ask the user to manually name every subskill.
Routing Table
| Task signal | Default mode | Route |
|---|---|---|
| development, feature, bugfix, multi-file implementation, pilot, full orchestration test | STRICT_OPENSPEC | OpenSpec bootstrap or bridge -> planning -> risk -> dispatch -> review -> QA -> delivery |
existing OpenSpec change path, openspec/changes, proposal/design/tasks, archive workflow | STRICT_OPENSPEC | bridge -> planning -> risk -> dispatch -> review -> QA -> delivery |
| development task where OpenSpec cannot be used | STRICT | planning -> risk -> dispatch -> review -> QA -> delivery |
| review-only request | LIGHT | review -> delivery |
| QA-only request | LIGHT | QA -> delivery |
| high-risk command, publishing, destructive write, credentials, external mutation | current mode plus risk gate | risk before execution |
If a task is tiny but still a development task, keep STRICT_OPENSPEC unless
the user explicitly asks for lightweight execution.
Strict Control Plane
For STRICT and STRICT_OPENSPEC, the run must use:
scripts/new_run.pyto initialize a run directoryledger.jsonas the execution state record- JSON
TaskManifest - JSON
WorkPackage - JSON or structured
AgentResult scripts/validate_contracts.pyfor core artifact validation- explicit review and QA gates
- final
control_surface_used
If a required script or artifact cannot be used, stop with a blocker instead of silently downgrading to markdown-only orchestration.
OpenSpec Bootstrap
When STRICT_OPENSPEC applies and no existing change is available, create or
request this scaffold:
openspec/changes/<change-id>/proposal.md
openspec/changes/<change-id>/design.md
openspec/changes/<change-id>/tasks.md
openspec/changes/<change-id>/ultra-bridge.md
Default to a single bounded change and one current implementation slice. OpenSpec owns durable specification state. Ultra owns execution control.
Slice-Driven Execution
Keep change and slice separate:
- OpenSpec
changeis the specification and progress ledger unit - Ultra
sliceis the implementation, verification, and commit unit
Use this slice status vocabulary:
slice_0_not_openedslice_0_spec_readyslice_1_completedslice_2_in_progressslice_3_qa_pendingslice_4_done
Every active change must name current slice status, next slice, owned paths, and the verification gate required before advancing.
State Machine
Default phases:
Intake -> Plan -> Dispatch -> Execute -> Review -> QA -> Deliver -> Retro
Required loopbacks:
Review -> Executewhen the plan is valid but the work is wrongQA -> Executewhen behavior is wrong but architecture is still validQA -> Planwhen the failure exposes a planning or requirement flaw
Default retry policy:
- transient failures: at most 1 bounded retry per work package
- repeated or hard blockers: escalate with summary, evidence, and next-step ask
Context Firewall
Pass artifacts, not full chat history:
- relevant
TaskManifest - assigned
WorkPackage - file pointers or artifact paths
- failure context for retries or reroutes
Safe Parallelism
Dispatch work in parallel only when:
- dependencies are satisfied in the DAG
owned_pathsdo not intersect with another active write package- inputs are stable enough to avoid speculative design
- acceptance checks are concrete enough for independent verification
If any condition is false, serialize.
Delivery Requirement
Every delivery must include:
final_deliverableorchestration_logvetter_reportcontrol_surface_used
control_surface_used must state:
run_modeused_openspec_changeused_openspec_bridgeused_run_ledgerused_contract_validationused_slice_dagused_dynamic_qa- skipped control surfaces and reasons
Supporting Skills
Route through these siblings as needed:
$openspec-ultra-bridgefor OpenSpec change mapping$decision-complete-plannerfor JSON-ready task graphs$dispatch-and-trackfor ledger and write-lock tracking$risk-vetterand$safety-guardfor risk gates and guardrails$spec-reviewand$code-reviewfor review gates$qa-verifyfor verification$deliver-and-retrofor final packaging
Read Next
- Read routing for strict routing examples.
- Read contracts for machine-checkable artifacts.
- Read state-machine for phase and loopback rules.
- Read design-tenets for governing principles.
What ships with it: 12 files
17.6 KB alongside SKILL.md, 3 of them executable
agents/
- openai.yaml337 B
references/
- contracts.md2.8 KB
- design-tenets.md772 B
- examples.md436 B
- review-lenses.md707 B
- routing.md1.9 KB
- source-synthesis.md1.3 KB
- state-machine.md675 B
- workflow.md1.5 KB
scripts/
- new_run.pyruns2.8 KB
- validate_contracts.pyruns2.6 KB
- validate_run.pyruns1.9 KB