OctoClaw Runtime

VerifiedSafe

TypeScript-first policy, delegation, status, and operator layer for OpenClaw.

Sby Skills Guide Bot
Data & AIAdvanced
7/22/2026
Claude CodeCursor
#multi-agent#orchestration#delegation#routing#cli

Recommended for

Automated summary

A TypeScript-first orchestration layer for OpenClaw that decides between direct reply and delegation, wraps delegated work in a WorkContract, and surfaces task state through dedicated tools.

Strengths

  • Cleanly separates the `reply` route (answer directly) from the `delegate` route, preventing unnecessary work creation
  • Requires every delegation to carry a traceable WorkContract: scope, allowed tools, route seal, and native spawn intent
  • Ships concrete CLI routing commands: config wizard, monthly budget, cost reports, and per-model score overrides or bans per complexity tier
  • Treats `task-state.json` as a rebuildable read-model cache instead of durable truth, with an explicit degraded-mode fallback

Limitations

  • Covers only the v0.6.0 line, flagged as unstable, and deliberately drops compatibility with the older Python runtime (`octoclaw_spawn`, completion files, JSON outbox)
  • Tightly coupled to the native OpenClaw runtime (TaskFlow, `sessions_spawn`, native announce) and its toolset, so it is unusable outside that ecosystem
  • Heavy jargon (route seal, spawn intent, status projection) makes onboarding slow
When to use it

Use it when an OpenClaw agent must decide between answering directly and delegating, then track or explain the state of work already delegated.

When not to use it

Avoid it on projects without the OpenClaw runtime, or when the goal is to recreate the retired Python octoclaw_spawn pipeline.

Security analysis

Safe
Quality score82/100

The skill is an internal runtime policy/usage document. It does not instruct destructive shell commands, network exfiltration, secret handling, obfuscated payloads, or safety-disabling behavior. Referenced tools are OctoClaw-specific and no allowed-tools are declared, so execution risk is minimal.

No concerns found

How this analysis is made

Examples

Delegate with confirmation
Délègue l'analyse des logs d'erreur du service paiement à un sous-agent via octoclaw_dispatch, puis confirme avec octoclaw_dispatch_confirm une fois le runId natif accepté.
Check delegated task status
Où en est la tâche déléguée pour la migration de la base ? Utilise octoclaw_status pour afficher la timeline et les artefacts, sans relancer de travail.
Configure the router budget
Configure le routeur OctoClaw avec `octoclawctl router wizard`, fixe un budget mensuel de 50 $ et marque le modèle gpt-mini comme dispreferred pour le tier de complexité élevée.

name: octoclaw metadata: author: openclaw-community version: "0.6.0" timestamp: "2026-05-09" stable: false tags:

  • multi-agent
  • scheduler
  • productivity description: > OctoClaw is the TypeScript-first policy, delegation, status, and operator layer for OpenClaw. Use it when a turn should be routed through reply/delegate policy, WorkContract-backed delegation, runtime status, IM delivery, or feedback-loop tooling. postInstall: | OctoClaw is installed. Current live route authority is reply | delegate. Use octoclaw_dispatch for delegated work and octoclaw_status for status.

OctoClaw Runtime Skill

This skill describes the current v0.6.0 runtime line. It intentionally does not describe the old Python octoclaw_spawn.py, completion-file finalizer, or JSON delivery-outbox runtime. Those were historical mechanisms and must not be reintroduced as compatibility paths.

Current Authority

| Layer | Authority | |------|-----------| | live route | reply | delegate only | | execution lifecycle | OpenClaw native TaskFlow / sessions_spawn / native announce | | OctoClaw metadata | SQLite runtime ledger: WorkContract, route seal, native refs, spawn intent, audit events | | status display | generated projection from metadata + native lifecycle | | task-state.json | rebuildable read-model cache, not truth | | replay log | observability and evaluation evidence, not runtime truth |

Main-Agent Rules

  1. For direct answers, use reply; do not create work.
  2. For delegated work, call octoclaw_dispatch; do not hand-write spawn args.
  3. After native spawn is accepted, confirm with octoclaw_dispatch_confirm.
  4. For status/provenance/follow-up questions, use octoclaw_status or octoclaw_task_action; do not dispatch new work just to explain old work.
  5. Do not call or recreate octoclaw_spawn.
  6. Do not use completion files, child-finalizer recovery, or JSON delivery outbox as the normal final-delivery mechanism.
  7. Do not treat task-state.json as durable truth. Missing or corrupt task-state must be rebuilt or reported degraded.
  8. Do not inject raw child transcripts into the parent context.

Tool Use

| Tool | Use | |------|-----| | octoclaw_route_hint | Ask the runtime for route guidance before model/tool escalation. | | octoclaw_policy_decide / octoclaw_route | Resolve policy and route facts. | | octoclaw_dispatch | Create a WorkContract-backed delegation plan and native spawn intent. | | octoclaw_dispatch_confirm | Bind accepted native runId / childSessionKey evidence back to OctoClaw metadata. | | octoclaw_task_action | Read task details, queue, artifacts, or supported task operations. | | octoclaw_status | Render status/details/queue/timeline from projection. |

Auto Router Commands

| Command | Use | |---------|-----| | octoclawctl router wizard [--incremental] [--config answers.json] | Create or update local router plan/budget/privacy/language/restricted/same-provider config. | | octoclawctl router promotion review [--input shadow.jsonl] | Evaluate local shadow samples and append promotion decisions. | | octoclawctl router decisions [--since 7d] | Inspect shadow-to-live promotion decisions. | | octoclawctl router cost report [--period 7d] | Review local cost grouped by model, complexity, and route, including prediction and budget status. | | octoclawctl router cost budget set --monthly <usd> | Store the local monthly router budget. | | octoclawctl router score override <model> <tier>=<score> | Override a model score for a complexity tier. | | octoclawctl router model mark <model> --dispreferred-for <tier> | Soft-avoid a model for a tier. | | octoclawctl router model ban <model> --for <tier> | Hard-exclude a model for a tier. |

Delegation Contract

Delegated work must carry:

  • WorkContract id and route seal.
  • scope and allowed tools.
  • role / worker-pool / model profile metadata.
  • native spawn intent id.
  • accepted native refs after confirm.
  • status projection derived from native lifecycle and OctoClaw metadata.

spawnExecuted=true requires accepted native run evidence. A child session key by itself is only a reference; it is not execution proof.

Status Questions

These are status-surface requests, not new work:

  • "OctoClaw status"
  • "task status"
  • "which task is running?"
  • "why did it not dispatch?"
  • "who handled the previous task?"
  • "show queue/details/timeline"

Answer through octoclaw_status or octoclaw_task_action. If no verifiable record exists, say that clearly and do not invent a task.

Current Docs

  • README.md / README.zh-CN.md: operator-facing overview.
  • docs/octoclaw-architecture-map-2026-05-09.md: detailed module graph.
  • docs/octoclaw-ts-rebuild-design-v2.md: architecture baseline.
  • docs/octoclaw-runtime-convergence-cleanup-plan-2026-05-07.md: runtime cleanup handoff.
  • docs/octoclaw-repo-debt-cleanup-plan-2026-05-09.md: archive/version/terminology cleanup.
  • docs/archive/: historical material only.

Forbidden Legacy Paths

Do not reintroduce these as product runtime behavior:

  • octoclaw_spawn as a public tool or alias.
  • plugin runtime.subagent.run() fallback.
  • fake detached runtime capability registration.
  • child completion files as normal final protocol.
  • child-finalizer recovery loop.
  • JSON delivery outbox for new planner/native runtime tasks.
  • resident runner / tmux / ClawTeam as required substrate.

From the file to a first result

The download contains exactly the content displayed on this page. This copy may differ from the latest version in the source repository.

View the source and supporting files

This download contains SKILL.md only. Any scripts, references, templates and dependencies are not included. Get the required files from the source, preserving their directory structure, before continuing.

1. Download the file

Download SKILL.md

From the root of a test project, save the file as .skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md, or use the command below. This folder is for inspecting the skill before installation.

Command for macOS/Linux or WSL: requires a POSIX shell and curl. On Windows without WSL, use the download button and create the folder manually.

mkdir -p '.skill-downloads/octoclaw-runtime-mrwakf5p' && curl --fail --show-error --location 'https://skills-guides.com/api/skills/octoclaw-runtime-mrwakf5p/download' --output '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md.part' && mv '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md.part' '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md'

2. Check prerequisites

Read SKILL.md before using it. If it references scripts or external services, inspect them and prepare the required access in your test environment.

No recognized prerequisites section in this file. This does not mean it is self-contained: libraries, tools and access requirements may be mentioned elsewhere in the content or source repository.

The platforms on this page are catalog indications. The paths below follow tool documentation; they do not prove this skill has been run in those tools. No minimum tool version is confirmed for this listing.

3. Load the skill

After inspection, copy the complete folder into your project using one of the commands below, then open a new tool session. The command stops if a folder with the same name already exists: compare the files before replacing it.

Claude Code
test -f '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md' && test -s '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md' && mkdir -p '.claude/skills' && test ! -e '.claude/skills/octoclaw' && cp -R '.skill-downloads/octoclaw-runtime-mrwakf5p' '.claude/skills/octoclaw'

Cursor
test -f '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md' && test -s '.skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md' && mkdir -p '.cursor/skills' && test ! -e '.cursor/skills/octoclaw' && cp -R '.skill-downloads/octoclaw-runtime-mrwakf5p' '.cursor/skills/octoclaw'

OpenSkills

OpenSkills must already be installed. The local install/read procedure is checked with OpenSkills 1.5.0, which requires Node.js ≥ 20.6.0. Catalog slugs are not resolved automatically.

Official OpenSkills installation
openskills install './.skill-downloads/octoclaw-runtime-mrwakf5p'

openskills read 'octoclaw-runtime-mrwakf5p'

Reading should display the installed file content and its base directory. It does not run the skill or install its dependencies.

For an agent that reads AGENTS.md, then sync the index. This command changes AGENTS.md: review the diff before using it.

openskills sync

Manual method or another tool: attach SKILL.md to a conversation, or open it in an agent that can read files. Include any required supporting resources. No universal folder is recognized by every tool.

4. Try a first task

Use a short, synthetic input: an API endpoint, a small document or an interface brief. In a new conversation, attach the file or give the agent its path, then adapt this request:

Read .skill-downloads/octoclaw-runtime-mrwakf5p/SKILL.md and apply its instructions to this example: [describe a test input]. Identify any missing files, dependencies or access. Produce the result requested by the skill without inventing data or running unreviewed scripts.

Expected result: output matching the skill instructions, or a precise list of missing prerequisites. A read confirmation alone does not validate a first use. If an external resource is unavailable, that path remains untested.

Related skills