agentsclimarketplace

Openspec ultra bridge

Skill keithhegit/ultra-orchestration/skills/openspec-ultra-bridge

Ultra Orchestration 是一套开源的 AI 工程编排技能体系。它将 OpenSpec 规格主线(Program → Milestone → Change → Slice)与 Ultra-Orchestrator 执行阶段(Intake → Plan → Dispatch → Execute → Review → QA → Deliver → Retro)结合,提升交付稳定性与可审计性。

Install
npx -y skills add keithhegit/ultra-orchestration --skill openspec-ultra-bridge

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing 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.

What its author says it does

Copied from the file, not written here

Bridge OpenSpec change artifacts into Ultra-Orchestrator execution artifacts. Use when a repo follows OpenSpec for proposal, design, tasks, and archive, but execution should remain controlled by Ultra-Orchestrator through TaskManifest, WorkPackage, review, QA, risk gates, and ledger-based orchestration.

SKILL.md

6.0 KB, as published. Nobody here has run it

OpenSpec Ultra Bridge

Treat OpenSpec as the spec source-of-truth and Ultra-Orchestrator as the execution control plane.

This bridge is part of the mainline strict control plane. It is normally invoked by $ultra-orchestrator; users should not need to call it manually.

Priority

Absorb OpenSpec advantages without weakening the existing Ultra-Orchestrator architecture.

Prioritize these four advantages first:

  1. Persistent spec assets
  2. Change-based workflow
  3. Archive discipline
  4. Spec-first task decomposition input

Do not let this bridge replace Ultra's state machine, review, QA, risk vetter, safety guard, or execution ledger ownership.

Purpose

Use this skill to translate between:

  • OpenSpec change artifacts
  • Ultra-Orchestrator execution artifacts

The bridge must not replace OpenSpec. It must map OpenSpec artifacts into Ultra execution objects and map Ultra review and QA feedback back into OpenSpec update suggestions.

Inputs

Expect one or more of these OpenSpec artifacts:

  • proposal.md
  • design.md
  • tasks.md
  • spec delta artifacts
  • change metadata such as change-id

Optionally consume these Ultra-side artifacts:

  • TaskManifest
  • WorkPackage
  • AgentResult
  • RunOutput
  • ExecutionLedger summary

Produce

Generate these Ultra-side artifacts when possible:

  • TaskManifest
  • WorkPackage
  • change-id <-> task-id mapping notes
  • archive gate readiness notes

Generate these OpenSpec-side feedback outputs when possible:

  • task refinement suggestions
  • design update suggestions
  • failure-context updates
  • archive blocking reasons

Mapping rules

  • Map change-id to TaskManifest.id.
  • Map proposal summary to goal.
  • Map design constraints to constraints.
  • Map OpenSpec task dependencies to dependencies.
  • Map task acceptance criteria or spec delta checks to success_criteria and acceptance_checks.
  • Map affected components or files to owned_paths.

Orchestration rules

  • Preserve OpenSpec as the spec source-of-truth.
  • Preserve Ultra-Orchestrator as the orchestration source-of-truth.
  • Do not rewrite OpenSpec artifacts into Ultra-native storage unless the user explicitly asks for that migration.
  • Prefer artifact-driven handoff over conversation-history handoff.
  • Treat archive as gated by Ultra review, QA, and risk decisions.
  • Use OpenSpec as a higher-quality frontend for planning, not as a replacement for dispatch, review, or QA.

Worktree guidance

When Git Worktree is used:

  • Prefer one OpenSpec change per worktree.
  • Record worktree path in run metadata.
  • Use owned-path rules and active write locks before parallel execution.

Pilot guidance

When validating this bridge on a real project such as /teammate:

  • Start with one change at a time.
  • Prefer one feature task, one bug-fix task, and one task that forces a review or QA loopback.
  • Judge success by bridge stability, not by feature count.

Change and slice contract

This bridge must preserve a strict division:

  • OpenSpec change = specification source-of-truth and long-lived progress node
  • Ultra slice = execution source-of-truth for implementation progress inside a change

Do not collapse them into one concept.

When bridging, always make the current slice explicit for the selected change.

Required change scaffold

For every newly opened change, initialize at minimum:

  • proposal.md
  • design.md
  • tasks.md
  • ultra-bridge.md

The initial bridge status should default to:

  • slice_0_spec_ready if the change is opened and ready for implementation
  • slice_0_not_opened only for program-level ledgers or unopened planned changes

Required ultra-bridge fields

Each ultra-bridge.md should include or imply:

  • change_id
  • milestone
  • status
  • task_manifest_focus
  • work_package_scope
  • next_slice

If a field is not known yet, write the narrowest truthful placeholder rather than omitting slice status.

Status synchronization

After each completed slice, the bridge should push status intent back into OpenSpec-facing artifacts:

  • update ultra-bridge.md status
  • update tasks.md task checks when implementation actually landed
  • suggest roadmap or current-status tally changes when opened/unopened counts changed

Do not treat conversation text as an acceptable substitute for status synchronization.

Strict-mode output

When the run mode is STRICT_OPENSPEC, produce or hand off:

  • JSON-ready TaskManifest
  • JSON-ready WorkPackage set
  • change-to-task mapping
  • current slice status
  • next slice
  • owned paths for the active slice
  • review and QA targets
  • archive or status-sync blockers

Batch initialization mode

When a program has many unopened planned changes, the bridge may be used to:

  1. enumerate all planned changes
  2. classify them as opened or unopened
  3. create missing change skeletons in batches
  4. stamp each opened change with an initial slice status

This is the preferred way to normalize a long-running program before large implementation waves.

Efficiency feedback loop

When a single change is used as a trial run, the bridge should also emit:

  • whether the change size was appropriate
  • whether the owned-path scope was narrow enough
  • what friction came from spec mapping versus implementation
  • what should be added back into the bridge workflow template

Use this to improve bridge stability over time rather than treating every change as isolated.

Read references on demand

  • references/mapping.md for detailed field mapping
  • references/workflow.md for end-to-end fused workflow
  • references/archive-gate.md for archive readiness conditions
  • references/adoption-priorities.md for what to absorb first from OpenSpec
  • references/teammate-pilot.md for pilot acceptance on /teammate

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.