/openclaw
Deterministic compiled graphs for OpenClaw (importer, runtime, MCP tools)
$ npx -y skills add sbhooley/ainativelang --skill openclaw --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/openclaw
Context preview
The summary Claude sees to decide when to auto-load this skill.
Deterministic compiled graphs for OpenClaw (importer, runtime, MCP tools)
SKILL.md
openclaw.SKILL.mdname: AINL
description: Deterministic compiled graphs for OpenClaw (importer, runtime, MCP tools)
install: ./install.sh
install_alt: pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw
commands:
- ainl import <source> — convert Markdown / ecosystem packs into `.ainl` graphs
- ainl compile <file.ainl> — compile and validate deterministic graph IR
- ainl run <file.ainl> — run compiled workflow (when your graph uses the runtime path)
- ~/.openclaw/bin/ainl-run <file.ainl> — compile-then-run wrapper after install-mcp --host openclaw (add ~/.openclaw/bin to PATH)
tags:
- ainl
- graphs
- mcp
- openclaw
metadata:
openclaw:
kind: skill
upstream: https://github.com/sbhooley/ainativelang
package_pypi: ainativelang
mcp_server_key: ainl
config_file: ~/.openclaw/openclaw.json
workspace_default: ~/.openclaw/workspace
onboarding_note: >-
OpenClaw expects `npm install` + `openclaw onboard` for the host CLI; this skill adds AINL (pip) + MCP + ainl-run.
install_channels:
- ClawHub (when listed)
- Manual copy to ~/.openclaw/skills or <workspace>/skillsAINL for OpenClaw
**AINL** is an AI-native language that compiles **deterministic graphs**. In [OpenClaw](https://openclaw.ai/), use it to import workflows, compile them, run them, and expose **`ainl-mcp`** tools to the agent via **`~/.openclaw/openclaw.json`**.
**Package:** `ainativelang` **v1.8.0** (this tree; **PyPI** after publish — **`docs/RELEASING.md`**). Optional extras: `[mcp]`, `[solana]`.
Agent instructions
Follow this skill when the user is on **OpenClaw** (or wants **deterministic `.ainl` graphs** with **import / compile / run / MCP**). Prefer **`./install.sh`** from a checked-out skill folder, or the manual **`pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw`** path (alias **`ainl install-openclaw`**). OpenClaw does **not** use `zeroclaw skills install <url>` — install via **ClawHub** (if available) or **copy this folder** into **`~/.openclaw/skills`** or **`<workspace>/skills`**. After bootstrap, suggest the **morning briefing** prompt example below when it fits.
When to use this skill
- The user wants **portable, compiled workflows** (`.ainl`) instead of ad-hoc prompts.
- They mention **import**, **compile**, **morning briefing**, **clawflows**, **ecosystem**, or **deterministic graphs**.
- They need **MCP** access to validation / compile / ecosystem listing (`ainl-mcp`).
Install (pick one)
1. **From this skill directory (after ClawHub or manual copy):** run `./install.sh` That optionally refreshes the **OpenClaw CLI** via npm, upgrades **`ainl[mcp]`**, and runs **`ainl install-mcp --host openclaw`** (MCP merge into **`openclaw.json`**, **`~/.openclaw/bin/ainl-run`** wrapper).
2. **Manual (no skill checkout):** `pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw`
Commands the user cares about
| Command | Purpose | |--------|---------| | `ainl import …` | Import Markdown or ecosystem sources into `.ainl` | | `ainl compile …` | Compile / validate graphs | | `ainl run …` | Execute a graph via the CLI runtime where applicable | | `~/.openclaw/bin/ainl-run <file.ainl>` | After `install-mcp --host openclaw`: wrapper that compiles then runs |
After install — prompt suggestion
Tell the user they can say in OpenClaw:
> Import the morning briefing using AINL.
(They should point the importer at their Markdown or pack; `ainl import` subcommands match their source type.)
MCP
Configure the host so **`mcp.servers.ainl`** in **`~/.openclaw/openclaw.json`** runs **`ainl-mcp`** as a stdio MCP server (see AINL docs: *OpenClaw integration* / *External orchestration*). **`ainl install-mcp --host openclaw`** merges that entry when missing.
References
- AINL: [github.com/sbhooley/ainativelang](https://github.com/sbhooley/ainativelang)
- **Gold standard (profiles, caps, cron, bootstrap, verification):** [docs/operations/OPENCLAW_AINL_GOLD_STANDARD.md](https://github.com/sbhooley/ainativelang/blob/main/docs/operations/OPENCLAW_AINL_GOLD_STANDARD.md) — `openclaw_ainl_gold_standard`
- **Host briefing — AINL v1.8.0 (what ships vs what OpenClaw must wire; doc key `openclaw_host_ainl_1_2_8`):** [docs/operations/OPENCLAW_HOST_AINL_1_2_8.md](https://github.com/sbhooley/ainativelang/blob/main/docs/operations/OPENCLAW_HOST_AINL_1_2_8.md) — `openclaw_host_ainl_1_2_8`
- OpenClaw: [openclaw.ai](https://openclaw.ai/)
- Package: **`ainativelang`** **v1.8.0** (this tree; **PyPI** after publish — **`docs/RELEASING.md`**)
Read more
name: AINL
description: Deterministic compiled graphs for OpenClaw (importer, runtime, MCP tools)
install: ./install.sh
install_alt: pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw
commands:
- ainl import <source> — convert Markdown / ecosystem packs into `.ainl` graphs
- ainl compile <file.ainl> — compile and validate deterministic graph IR
- ainl run <file.ainl> — run compiled workflow (when your graph uses the runtime path)
- ~/.openclaw/bin/ainl-run <file.ainl> — compile-then-run wrapper after install-mcp --host openclaw (add ~/.openclaw/bin to PATH)
tags:
- ainl
- graphs
- mcp
- openclaw
metadata:
openclaw:
kind: skill
upstream: https://github.com/sbhooley/ainativelang
package_pypi: ainativelang
mcp_server_key: ainl
config_file: ~/.openclaw/openclaw.json
workspace_default: ~/.openclaw/workspace
onboarding_note: >-
OpenClaw expects `npm install` + `openclaw onboard` for the host CLI; this skill adds AINL (pip) + MCP + ainl-run.
install_channels:
- ClawHub (when listed)
- Manual copy to ~/.openclaw/skills or <workspace>/skillsAINL for OpenClaw
**AINL** is an AI-native language that compiles **deterministic graphs**. In [OpenClaw](https://openclaw.ai/), use it to import workflows, compile them, run them, and expose **`ainl-mcp`** tools to the agent via **`~/.openclaw/openclaw.json`**.
**Package:** `ainativelang` **v1.8.0** (this tree; **PyPI** after publish — **`docs/RELEASING.md`**). Optional extras: `[mcp]`, `[solana]`.
Agent instructions
Follow this skill when the user is on **OpenClaw** (or wants **deterministic `.ainl` graphs** with **import / compile / run / MCP**). Prefer **`./install.sh`** from a checked-out skill folder, or the manual **`pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw`** path (alias **`ainl install-openclaw`**). OpenClaw does **not** use `zeroclaw skills install <url>` — install via **ClawHub** (if available) or **copy this folder** into **`~/.openclaw/skills`** or **`<workspace>/skills`**. After bootstrap, suggest the **morning briefing** prompt example below when it fits.
When to use this skill
- The user wants **portable, compiled workflows** (`.ainl`) instead of ad-hoc prompts.
- They mention **import**, **compile**, **morning briefing**, **clawflows**, **ecosystem**, or **deterministic graphs**.
- They need **MCP** access to validation / compile / ecosystem listing (`ainl-mcp`).
Install (pick one)
1. **From this skill directory (after ClawHub or manual copy):** run `./install.sh` That optionally refreshes the **OpenClaw CLI** via npm, upgrades **`ainl[mcp]`**, and runs **`ainl install-mcp --host openclaw`** (MCP merge into **`openclaw.json`**, **`~/.openclaw/bin/ainl-run`** wrapper).
2. **Manual (no skill checkout):** `pip install 'ainativelang[mcp]' && ainl install-mcp --host openclaw`
Commands the user cares about
| Command | Purpose | |--------|---------| | `ainl import …` | Import Markdown or ecosystem sources into `.ainl` | | `ainl compile …` | Compile / validate graphs | | `ainl run …` | Execute a graph via the CLI runtime where applicable | | `~/.openclaw/bin/ainl-run <file.ainl>` | After `install-mcp --host openclaw`: wrapper that compiles then runs |
After install — prompt suggestion
Tell the user they can say in OpenClaw:
> Import the morning briefing using AINL.
(They should point the importer at their Markdown or pack; `ainl import` subcommands match their source type.)
MCP
Configure the host so **`mcp.servers.ainl`** in **`~/.openclaw/openclaw.json`** runs **`ainl-mcp`** as a stdio MCP server (see AINL docs: *OpenClaw integration* / *External orchestration*). **`ainl install-mcp --host openclaw`** merges that entry when missing.
References
- AINL: [github.com/sbhooley/ainativelang](https://github.com/sbhooley/ainativelang)
- **Gold standard (profiles, caps, cron, bootstrap, verification):** [docs/operations/OPENCLAW_AINL_GOLD_STANDARD.md](https://github.com/sbhooley/ainativelang/blob/main/docs/operations/OPENCLAW_AINL_GOLD_STANDARD.md) — `openclaw_ainl_gold_standard`
- **Host briefing — AINL v1.8.0 (what ships vs what OpenClaw must wire; doc key `openclaw_host_ainl_1_2_8`):** [docs/operations/OPENCLAW_HOST_AINL_1_2_8.md](https://github.com/sbhooley/ainativelang/blob/main/docs/operations/OPENCLAW_HOST_AINL_1_2_8.md) — `openclaw_host_ainl_1_2_8`
- OpenClaw: [openclaw.ai](https://openclaw.ai/)
- Package: **`ainativelang`** **v1.8.0** (this tree; **PyPI** after publish — **`docs/RELEASING.md`**)
Find Us on X: @ainativelang AI-led co-development project, human-initiated by Steven Hooley (x.com/sbhooley, stevenhooley.com, linkedin.com/in/sbhooley). Attribution details: docs/PROJECT_ORIGIN_AND_ATTRIBUTION.md and tooling/project_provenance.json.

