返回 Skill 列表
extension
分类: 其它无需 API Key

Mova Spec Guide

回答关于 MOVA 规范语言的问题——包括模式(ds.*)、信封(env.*)、动词、剧集、全局目录、指令配置文件等。

person作者: mova-compacthubclawhub

Ecosystem Skill — Supports building and managing the MOVA ecosystem. Requires the openclaw-mova plugin.

MOVA Spec Guide

Answer any question about the MOVA language — from basic concepts to schema-level validation — by reading the canonical spec documents in the workspace and referencing the public repository.

What this skill does

  1. Explains MOVA concepts — schemas (ds.*), envelopes (env.*), verbs, episodes, global catalogs, instruction profiles, runtime bindings, connectors, security layer, text/UI layer
  2. Reads authoritative source docs — uses the local spec clone for immediate access; references the public GitHub repo for the canonical latest state
  3. Shows examples — points to examples/ files and explains them in context
  4. Validates artifacts — given a JSON draft, checks it against the relevant MOVA schema and explains violations
  5. Maps concepts to practice — explains how a concept applies to a real use case the user brings

Source hierarchy

| Priority | Source | Use for | |----------|--------|---------| | 1 (primary) | Local workspace: /home/mova/.openclaw/workspace/mova-spec/ | All reads — fast and always available | | 2 (canonical reference) | GitHub: https://github.com/mova-compact/mova-spec | Latest version check, stable URLs to share |

Always read from the local workspace first. Only reference the GitHub repo if the user needs a link to share or asks about the latest uncommitted changes.

Workspace layout

mova-spec/
  docs/
    mova_core.md                      ← core specification (start here for concept questions)
    mova_global_and_verbs.md          ← verbs, global catalogs, action_signature
    mova_operator_frame.md            ← operator frame: 13-axis audit lens
    mova_episodes_and_genetic_layer.md← episode structure and pattern memory
    mova_layers_and_namespaces.md     ← red-core / skills / infra layering
    mova_runtime_and_connectors.md    ← runtime bindings and connector contracts
    mova_security_layer.md            ← instruction profiles and security events
    mova_text_and_ui_layer.md         ← text channel separation
    MOVA_6.0.0_RELEASE_NOTES.md       ← latest changes: verb/tool/action_signature
  schemas/
    ds.mova_schema_core_v1.schema.json
    ds.mova_episode_core_v1.schema.json
    ds.instruction_profile_core_v1.schema.json
    ds.connector_core_v1.schema.json
    ds.runtime_binding_core_v1.schema.json
    ds.ui_text_bundle_core_v1.schema.json
    ds.mova4_core_catalog_v1.schema.json
    ds.security_event_episode_core_v1.schema.json
    env.*.schema.json (envelopes)
  examples/
    mova4_core_catalog.example.json
    action_signature.example.json
    env.*.example.json
  global.episode_type_catalog_v1.json
  global.security_catalog_v1.json
  global.layers_and_namespaces_v1.json
  global.text_channel_catalog_v1.json
  README-mova-spec.md

Core MOVA concepts (quick map)

| Concept | What it is | Where defined | |---------|-----------|---------------| | ds.* schema | JSON Schema describing what data looks like | schemas/ds.*.schema.json | | env.* envelope | Typed speech-act: a request, command, or event over ds.* data | schemas/env.*.schema.json | | Verb | Abstract operation type: create, update, route, record, publish, analyze, ... | docs/mova_global_and_verbs.md | | Tool | Execution channel (tool_id); 0 = tool-less | docs/mova_global_and_verbs.md §4.5 | | Action signature | Atomic unit for policy: (verb_id, tool_id, target_kind?) | MOVA_6.0.0_RELEASE_NOTES.md | | Episode | Structured record of one meaningful work step | ds.mova_episode_core_v1 | | Instruction profile | Declarative policy and guardrail set for an executor | ds.instruction_profile_core_v1 | | Global catalog | Shared vocabulary for all layers | global.*.json | | Operator frame | 13-axis audit lens (what/how/where/when/why/...) | docs/mova_operator_frame.md | | Genetic layer | Pattern memory built from episodes over time | docs/mova_episodes_and_genetic_layer.md |

How to handle different question types

Concept question

Read the relevant docs/ file. Summarize the answer. Quote the exact definition. Show a minimal example if one exists in examples/.

Schema question

Read the relevant schemas/*.schema.json. Explain each field: type, required/optional, allowed values. Show how the field relates to the concept.

Validation question

Read the user's JSON. Read the target schema. Check required fields, field types, allowed enum values. Report each violation with the relevant schema field and rule.

How to write question

Read the most relevant docs/ file. Walk through a minimal valid instance step by step. Reference the examples/ file if it exists.

Catalog lookup

Read the relevant global.*.json. Quote the exact entry. Explain the purpose in one sentence.

MOVA version

Current version in the workspace: MOVA 6.0.0

Key additions in 6.0.0 (see MOVA_6.0.0_RELEASE_NOTES.md):

  • Normative definition of verb, tool, action_signature
  • tool_id = 0 is canonical for tool-less actions
  • Policy matching priority: action_signature > verb_id > tool_id
  • action_labels in domain dictionaries (readability, no normative weight)

Rules

  • NEVER invent schema field names, verb IDs, or catalog values — always read from workspace files
  • NEVER reference schema URLs without first reading the local file to confirm the content
  • If the user asks about a concept not yet defined in the spec — say so explicitly, do not improvise
  • Always cite which document and section the answer comes from
  • If the user provides a JSON artifact to validate — read the correct schema file first, then compare field by field
  • MOVA does not execute — never describe MOVA artifacts as having runtime behavior