44 lines (30 loc) · 2.02 KB

ActiveClaw Workspace Contract

Purpose: make repo entry, verification, and mutable-state boundaries source-visible so future collaborators do not depend on thread memory.

What this repo is

ActiveClaw runtime: sessions, tools, memory, channel integrations, cron execution, and agent operating surface.
Canonical operator docs + source should answer behavior questions before thread memory does.

Canonical paths

Core docs root: docs/
Agent skills: skills/
CLI/runtime source: src/
Package/runtime metadata: package.json
Local OpenClaw config for this workspace lives outside the repo at ~/.openclaw/openclaw.json
Workspace directives live outside the repo at ~/.openclaw/workspace/

Runtime-data boundary

Treat this repo as code, docs, and checked-in operational logic.
Treat ~/.openclaw/ as mutable runtime state (logs, config, sessions, cron state, memory indexes, channel state).
Do not infer current behavior from old thread memory when local source, docs, config, or live runtime state can answer it.

Verification entrypoints

For OpenClaw behavior/commands/config: read local docs in docs/ first.
For actual runtime settings: verify ~/.openclaw/openclaw.json.
For live status/behavior questions: prefer source + runtime inspection over conversational recall.
If a claim is about a tool surface, command, config field, or agent behavior, verify the local source file before repeating it.

Local source wins

If thread memory, docs, or recollection conflict with local source/config/runtime artifacts, local source wins.
Prefer exact files over memory for field names, endpoints, command shapes, and config semantics.
Ship small checked-in contracts/docs when a repo boundary keeps causing repeated orientation drift.

Smallest useful collaboration default

When handing work to another agent, point them to:

WORKSPACE.md for repo contract
the exact source/doc path that defines the behavior in question
any external mutable state path they must verify before making claims