Spec-first target lifecycle manager for OpenClaw agents, skills, and pipelines.
- Interactive interviews — guided create/adopt/extend workflows for agents, skills, and pipelines
- Spec-first paradigm — canonical YAML specs define targets; runtime artifacts are compiled output
- 8 audit checks — automated readiness assessment before deployment
- Pipeline registry — track managed targets with lifecycle status
- Dual-format skills — 7 SKILL.md files for both OpenClaw and Claude Code
- Plugin hooks — abstract interfaces for governance, outbound gate, and MCP integrations
pip install clawscaffoldWith optional spec validation:
pip install clawscaffold[spec]# Initialize a project
clawscaffold init
# Create a new agent spec
clawscaffold create --name sales/coach --kind agent
# Adopt an existing agent
clawscaffold adopt --name sales/coach --source agents/sales/coach/SOUL.md --kind agent
# Audit a target
clawscaffold audit --name sales/coach
# Render spec to runtime files
clawscaffold render --id sales/coach# For Claude Code
./scripts/install-skills.sh claude-code
# For OpenClaw
./scripts/install-skills.sh openclaw| Command | Description |
|---|---|
init |
Initialize a clawscaffold project (creates .clawscaffold marker) |
create |
Create a new target spec |
adopt |
Adopt an existing runtime artifact into managed scope |
audit |
Run readiness checks against a target |
interview |
Start an interactive interview for target creation/adoption |
render |
Compile spec to runtime files |
validate |
Validate spec structure and content |
promote |
Promote a target to managed status |
enforce |
Enforce managed registry consistency |
install-skills |
Install SKILL.md files to Claude Code or OpenClaw |
sync-paperclip |
Sync agents to Paperclip for org chart and task management |
ClawScaffold includes an optional adapter for Paperclip that bulk-registers agents with hierarchy, generates API keys, and keeps agent instructions in sync.
The adapter activates only when Paperclip is detected (PAPERCLIP_API_URL env var or pnpm paperclipai CLI available). No Python Paperclip dependencies are added.
# Preview what would be imported (no changes made)
clawscaffold sync-paperclip --dry-run
# Bulk-register all catalog agents to Paperclip
PAPERCLIP_COMPANY_ID=<uuid> clawscaffold sync-paperclip
# Filter to a department
clawscaffold sync-paperclip --filter "executive/*"
# Register and generate API keys
clawscaffold sync-paperclip --generate-keys
# Sync a single agent after create/adopt
clawscaffold create --kind agent --id sales/coach --sync-paperclipWhat it does:
- Generates
.paperclip.yamlportable format fromcatalog/agents/specs - Converts
SOUL.mdfiles to PaperclipAGENTS.mdwith frontmatter - Shells out to
pnpm paperclipai company importfor bulk registration - Generates API keys via
pnpm paperclipai agent local-cli - Stores keys at
~/.openclaw/workspace/paperclip-agent-keys/<key>.json
ClawSpec is the contract-first QA framework for OpenClaw. When used together, ClawScaffold auto-generates ClawSpec test scenarios during adoption and creation workflows.
Install both:
pip install clawscaffold[spec] # includes clawspec as optional dependencyWhen ClawSpec is installed, clawscaffold adopt and clawscaffold create automatically generate:
tests/scenarios.yaml— smoke, negative, and identity test contractstests/handoffs/*.yaml— handoff contracts for multi-agent delegation- Coverage ledger entries for the adopted target
Without ClawSpec installed, these steps are skipped gracefully — the scaffolder works standalone for spec management and auditing.
# Adopt an agent — ClawSpec scenarios auto-generated
clawscaffold adopt --name sales/coach --source agents/sales/coach/SOUL.md --kind agent
# Verify the generated tests
clawspec validate agents/sales/coach
clawspec run agents/sales/coach --dry-run- ClawSpec — Contract-first QA for OpenClaw skills and agents (29 assertion types, Opik observability, regression baselines)
- ClawWrap — Spec-first outbound message routing gate (policy enforcement, audit trail)
- OpenClaw — Local-first AI agent framework (LLM + chat channels + Markdown skills)
- Paperclip — AI agent orchestration platform (org chart, task management, conversations)
git clone https://github.com/austinmao/clawscaffold.git
cd clawscaffold
pip install -e ".[dev]"
pytestClawScaffold ships with a gateway plugin (scaffold-planner) that registers scaffold interview tools directly in the OpenClaw gateway. The plugin provides four tools: scaffold_analyze, scaffold_next_question, scaffold_answer, and scaffold_finalize.
- Copy the
extensions/scaffold-planner/directory into your OpenClaw workspace:
cp -r extensions/scaffold-planner/ ~/.openclaw/extensions/scaffold-planner/- Register the plugin in your
~/.openclaw/openclaw.json:
{
"extensions": {
"scaffold-planner": {
"repoRoot": "/path/to/your/workspace",
"timeoutMs": 30000
}
}
}- Restart the gateway:
openclaw gateway restart| Option | Type | Default | Description |
|---|---|---|---|
repoRoot |
string | $OPENCLAW_WORKSPACE or cwd |
Workspace root containing the clawscaffold package |
pythonBin |
string | python3 |
Path to Python binary |
timeoutMs |
integer | 30000 |
CLI execution timeout in milliseconds |
scaffold_analyze, scaffold_next_question, scaffold_answer, scaffold_finalize
See the openclaw.plugin.json for the full config schema.
This package is part of ClawSuite — the OpenClaw agent infrastructure toolkit.
| Package | Description | Repo |
|---|---|---|
| ClawPipe | Config-driven pipeline orchestration | austinmao/clawpipe |
| ClawSpec | Contract-first testing for skills & agents | austinmao/clawspec |
| ClawWrap | Outbound policy & conformance engine | austinmao/clawwrap |
| ClawAgentSkill | Skill discovery, scanning & adoption | austinmao/clawagentskill |
| ClawScaffold | Agent/skill scaffold interviews | austinmao/clawscaffold |
| ClawInterview | Pipeline interview compilation & execution | (coming soon) |
All packages include OpenClaw gateway plugins for autonomous agent access.
MIT