⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
project-orchestrator
>
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
Project Orchestrator
Structured project lifecycle orchestration for OpenClaw.
This skill adds a deterministic state machine around tracked software projects so work moves through explicit stages, required artifacts, review loops, and approval gates instead of ad-hoc status changes.
What it is for
Use this skill when you want OpenClaw to manage a real project in a workspace that already tracks projects in PROJECTS.yaml.
It is most useful when you want:
- predictable stage transitions
- explicit approval gates before major work proceeds
- artifact validation before transitions
- structured producer/critic review loops
- optional Linear project and issue sync
- a clear paper trail in
projects/<name>.md
It is not a general task manager for one-off questions or untracked work.
Lifecycle
INTAKE -> BRIEF -> [DESIGN] -> [ARCHITECTURE] -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
Rules:
DESIGNis optional and intended for UI-heavy work.ARCHITECTUREis only used forprojecttier work.- Any state can transition to
CANCELED. - Approval gates exist at
BRIEF,ARCHITECTURE,PLAN, andREVIEW.
Tier flows
patch
INTAKE -> BRIEF -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
feature
INTAKE -> BRIEF -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
feature with UI design
INTAKE -> BRIEF -> DESIGN -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
project
INTAKE -> BRIEF -> ARCHITECTURE -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
project with UI design
INTAKE -> BRIEF -> DESIGN -> ARCHITECTURE -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED
Workspace expectations
This skill assumes a standard OpenClaw workspace with at least:
PROJECTS.yamlprojects/skills/project-orchestrator/
The scripts auto-detect the workspace by:
OPENCLAW_WORKSPACE- walking upward from the current directory looking for
PROJECTS.yaml - falling back to
~/.openclaw/workspace
Prerequisites
Required
- Python 3.10+
- an OpenClaw workspace using
PROJECTS.yaml
Optional
- Linear credentials if you want project and issue sync
anthropicPython package plus an Anthropic API key if you want to use the design generator- a runtime that can launch subagents if you want the full producer/critic review workflow
Install
From your workspace root:
mkdir -p ~/.openclaw/workspace/skills
cp -r /path/to/openclaw-agent-skills/project-orchestrator ~/.openclaw/workspace/skills/
Or symlink it:
ln -s /path/to/openclaw-agent-skills/project-orchestrator ~/.openclaw/workspace/skills/project-orchestrator
Core commands
Run these from the workspace root.
python3 skills/project-orchestrator/scripts/orchestrator.py init <name> --tier <patch|feature|project>
python3 skills/project-orchestrator/scripts/orchestrator.py status <name>
python3 skills/project-orchestrator/scripts/orchestrator.py validate <name>
python3 skills/project-orchestrator/scripts/orchestrator.py transition <name> <TARGET_STATE>
python3 skills/project-orchestrator/scripts/orchestrator.py plan <name>
python3 skills/project-orchestrator/scripts/orchestrator.py review-status <name>
All commands return JSON.
Typical workflow
- Initialize a tracked project with
init. - Fill in the project entry in
PROJECTS.yaml. - Write the current-stage artifact in
projects/<name>.mdusing the provided templates. - Run
validatebefore every approval gate or transition. - Use
transitionto move to the next valid state. - Keep review files and audit artifacts alongside the project summary.
Review model
The recommended pattern at approval gates is:
producer -> critic -> project-manager review -> operator approval
In practice that means:
- one agent or subagent produces the artifact
- a second agent critiques it and raises issues
- a coordinating project-manager pass checks whether the result is ready
- the human operator gives final approval
The repository includes a reusable review template at:
references/templates/inter-agent-review.md
PM relay watchdog, correct usage
When your runtime cannot keep a persistent PM session alive, use the PM relay helper as an explicitly scoped watchdog.
Recommended commands:
python3 skills/project-orchestrator/scripts/pm-relay-helper.py activate <name> --state <STATE>
python3 skills/project-orchestrator/scripts/pm-relay-helper.py list-active
python3 skills/project-orchestrator/scripts/pm-relay-helper.py sweep-active
python3 skills/project-orchestrator/scripts/pm-relay-helper.py deactivate <name>
Rules:
- track only projects that are actively in flight
- keep the active-project list as the single source of truth for relay scope
- do not run global sweeps across every tracked project
- do not respawn a PM owner if a real blocker already explains the gap
- do not respawn into an operator-approval-only wait state
- keep idle watchdog runs silent, and only send user-facing updates when the watchdog actually changes something meaningful
Linear integration
Linear support is built in, but it is a workflow choice rather than a documentation prerequisite.
Relevant helpers:
python3 skills/project-orchestrator/scripts/linear_integration.py create-project
python3 skills/project-orchestrator/scripts/linear_integration.py create-issue
python3 skills/project-orchestrator/scripts/linear_integration.py create-issues-from-plan
python3 skills/project-orchestrator/scripts/linear_integration.py update-state
python3 skills/project-orchestrator/scripts/linear_integration.py sync-state
Set the required Linear environment variables in the workspace where you run the scripts.
Optional design workflow
For UI-heavy projects, the skill includes a design generator and SVG wireframe generator.
python3 skills/project-orchestrator/scripts/design-producer.py --brief <file> --output-dir <dir> --wireframes
python3 skills/project-orchestrator/scripts/wireframe-gen.py --input <spec.json> --output-dir <dir>
The design generator looks for an API key in this order:
--api-keyANTHROPIC_API_KEY- an OpenClaw
auth-profiles.jsonif present
Included files
project-orchestrator/
├── SKILL.md
├── README.md
├── scripts/
│ ├── orchestrator.py
│ ├── linear_integration.py
│ ├── pm-checker.py
│ ├── design-producer.py
│ └── wireframe-gen.py
├── references/
│ ├── state-machine.yaml
│ ├── design-workflow.md
│ └── templates/
└── tests/
Notes on strictness
This skill is intentionally opinionated.
It favors:
- explicit states over informal progress updates
- validation over trust
- review loops over single-pass artifacts
- operator approval over silent transitions
If that is the operating model you want, this skill is a good fit.
// HOW IT'S BUILT
// FILES
2 files shipped with this skill. Inline text is the catalog copy. The upstream link is the source repository.
project-orchestrator/SKILL.md7.8 KBupstream ↗
--- name: project-orchestrator description: > Orchestrate tracked software projects with a deterministic lifecycle state machine, required artifacts, approval gates, review loops, and optional Linear synchronization. Use when: creating a new tracked project, checking project status, validating project artifacts, transitioning project state, planning project work, or enforcing a stage-based workflow for an entry in PROJECTS.yaml. NOT for: ad-hoc tasks, one-off questions, or work that is not being tracked as a project in the current OpenClaw workspace. --- # Project Orchestrator Use this skill to manage tracked software projects through explicit states. The scripts validate states, transitions, and required artifacts against `references/state-machine.yaml`. ## Workspace assumptions Run from an OpenClaw workspace that uses `PROJECTS.yaml` and `projects/`. Workspace detection order: 1. `OPENCLAW_WORKSPACE` 2. walk upward from the current directory looking for `PROJECTS.yaml` 3. fallback to `~/.openclaw/workspace` ## Quick reference All commands output JSON. ```bash python3 skills/project-orchestrator/scripts/orchestrator.py init <name> --tier <patch|feature|project> python3 skills/project-orchestrator/scripts/orchestrator.py status <name> python3 skills/project-orchestrator/scripts/orchestrator.py validate <name> python3 skills/project-orchestrator/scripts/orchestrator.py transition <name> <TARGET_STATE> python3 skills/project-orchestrator/scripts/orchestrator.py plan <name> python3 skills/project-orchestrator/scripts/orchestrator.py review-status <name> ``` If you are already in the skill directory, `python3 scripts/orchestrator.py ...` also works. ## Tiers Choose the tier at `INTAKE` based on scope. | Tier | Token budget | Typical scope | Uses `ARCHITECTURE`? | |------|--------------|---------------|----------------------| | patch | < 50k | Small, local change | No | | feature | 50k-300k | Contained feature in an existing system | No | | project | > 300k | New repo or major architectural change | Yes | Estimate in tokens, not time. When unsure, start with `feature`. Upgrade to `project` if major architecture work emerges during `BRIEF`. ## State flow **patch / feature** ```text INTAKE -> BRIEF -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED ``` **project** ```text INTAKE -> BRIEF -> ARCHITECTURE -> PLAN -> BUILD -> REVIEW -> SHIP -> CLOSED ``` **optional UI design path** ```text INTAKE -> BRIEF -> DESIGN -> PLAN/ARCHITECTURE -> ... ``` Any state can transition to `CANCELED`. ## Approval gates Approval gates exist at `BRIEF`, `ARCHITECTURE`, `PLAN`, and `REVIEW`. At each gate: 1. Run `validate <name>`. 2. Confirm the required artifact and review files exist. 3. Present a concise summary to the human operator in the current working channel. 4. Wait for explicit approval. 5. Only then run `transition <name> <TARGET_STATE>`. Never auto-approve. Never skip approval gates. ## New project workflow 1. Run `init <name> --tier <tier>`. 2. Update the project entry in `PROJECTS.yaml` with repo, path, description, tags, and any external IDs. 3. Write the current-stage artifact in `projects/<name>.md`. 4. Use the templates in `references/templates/`. 5. Run `transition <name> BRIEF` once intake is complete. 6. Continue state by state, validating before every transition. ## Status workflow 1. Run `status <name>`. 2. Read the current state, exit criteria, required artifacts, and valid transitions. 3. Report the result in plain language. ## Transition workflow 1. Run `validate <name>` first. 2. If validation fails, fix the missing artifact or review requirement before proceeding. 3. If the current state is an approval gate, get explicit human approval. 4. Run `transition <name> <TARGET_STATE>`. 5. Confirm the returned JSON shows the expected new state and follow-up actions. 6. If your workspace uses Linear, verify the sync fields in the response and correct failures immediately. ## PM relay watchdog, correct usage When your runtime cannot keep a persistent PM session alive, use the PM relay helper as an explicitly scoped watchdog. Recommended commands: ```bash python3 skills/project-orchestrator/scripts/pm-relay-helper.py activate <name> --state <STATE> python3 skills/project-orchestrator/scripts/pm-relay-helper.py list-active python3 skills/project-orchestrator/scripts/pm-relay-helper.py sweep-active python3 skills/project-orchestrator/scripts/pm-relay-helper.py deactivate <name> ``` Rules: - track only projects that are actively in flight - keep the active-project list as the source of truth for relay scope - do not run global sweeps across every tracked project - do not respawn a PM owner if a real blocker already explains the gap - do not respawn into an operator-approval-only wait state - keep idle watchdog runs silent, and only surface user-facing updates when the watchdog actually changes something meaningful ## Linear integration Linear support is built into the scripts. Use it when your workspace tracks projects there. ### Commands ```bash python3 skills/project-orchestrator/scripts/linear_integration.py create-project python3 skills/project-orchestrator/scripts/linear_integration.py create-issue python3 skills/project-orchestrator/scripts/linear_integration.py create-issues-from-plan python3 skills/project-orchestrator/scripts/linear_integration.py update-state python3 skills/project-orchestrator/scripts/linear_integration.py sync-state python3 skills/project-orchestrator/scripts/linear_integration.py add-comment python3 skills/project-orchestrator/scripts/linear_integration.py get-issue python3 skills/project-orchestrator/scripts/linear_integration.py validate-transition ``` ### Rules - Keep orchestrator state and Linear state aligned if the project has a `linear_project_id`. - Use your workspace's configured bot or service account for agent-owned issues. - Assign human-only tasks to the human operator. - When work is completed during `BUILD`, update the related Linear issue state immediately. ### State mapping | Orchestrator state | Linear state | |--------------------|--------------| | INTAKE, BRIEF | Backlog | | ARCHITECTURE | Todo | | PLAN | In Progress | | BUILD | In Dev | | REVIEW | Review | | SHIP | In Prod | | CLOSED | Done | | CANCELED | Canceled | ## Templates Templates live in `references/templates/`. - `brief.md` - `BRIEF` - `design-spec.md` - `DESIGN` - `architecture.md` - `ARCHITECTURE` - `plan.md` - `PLAN` - `review-checklist.md` - `REVIEW` - `inter-agent-review.md` - approval-gate review loops ## Inter-agent review workflow Approval-gate artifacts should go through a review loop before operator approval. Pattern: ```text producer -> critic -> project-manager review -> operator approval ``` ### Recommended roles by stage | Stage | Producer | Critic | |-------|----------|--------| | BRIEF | main agent or project lead agent | domain expert reviewer | | ARCHITECTURE | architect agent | senior engineer reviewer | | PLAN | planner agent | tech lead reviewer | | REVIEW | reviewer agent | QA reviewer | ### Review process 1. Producer writes or updates the artifact. 2. Critic reviews it and records issues as `BLOCKING`, `SIGNIFICANT`, or `MINOR`. 3. Producer responds and revises. 4. Repeat up to 3 rounds. 5. A project-manager pass decides whether the artifact is ready for operator review. 6. Present the artifact and review summary to the operator. ### Review files Store review files in `projects/` and use the template at `references/templates/inter-agent-review.md`. Suggested naming: ```text projects/<name>-review-<stage>-round<N>.md ``` Run: ```bash python3 skills/project-orchestrator/scripts/orchestrator.py review-status <name> ``` Use the result to confirm whether review files exist and whether sign-off is complete. ## Pre-close verifications Before moving from `SHIP` to `CLOSED`, require two independent checks. ### 1. Test verification Have an independent reviewer inspect tREADME.mdupstream ↗
Contents are not inlined in the catalog. Open the upstream file.
Plain-text copy, no login: /skill-bundles/justinstolzenberg-sudo-openclaw-agent-skills-project-orchestrator.md