CategoryWeightCategory scoreContribution
Securityprompt, exec, net, exfil, eval
35%
30.8 pts
Supply chainhash, typosquat, maintainer, lockfile
20%
20.0 pts
Maintenancestaleness, pinning, CI
15%
15.0 pts
TransparencySKILL.md, perms, README
15%
15.0 pts
Communityinstalls, verify, response
15%
15.0 pts
Findings & checks · 1 flagged
Security
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · _workspace/cycle-briefing-1.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
The exact value spotted
103## Notes
104
105- Auto mode ON: agents do not ask for confirmation on operational details. A real decision o
… (39 chars elided on L105)
106- Any agent that needs to change a contract: pause, write a note in `_workspace/contracts/PR
… (71 chars elided on L106)
107- Work goes directly into `live/` and root/`src/` (not drafted in `_workspace/`) because it
… (52 chars elided on L107)
Occurrences
1 occurrence · at L105How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
- Delete blanket "don't ask / no need to confirm" directives from the skill.
- If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051Trace & refs
Supply chain
✓—All supply chain checks passedNo findings in this category for the latest scan.pass
Maintenance
✓—All maintenance checks passedNo findings in this category for the latest scan.pass
Transparency
✓—All transparency checks passedNo findings in this category for the latest scan.pass
Community
✓—All community checks passedNo findings in this category for the latest scan.pass
Every scanned point with the score it earned and what moved between them.
80688be→80688be100 → 96-4
Score fell 4 points between these scans.
Computing diff…
The primary manifest — the file an agent reads to learn what this artifact does.
ableton-mind
Definitive MCP (Model Context Protocol) server for Ableton Live. Exposes the full Live Object Model to LLMs (Claude, Cursor, etc.) with an embedded native device knowledge base, declarative music recipes, an integrated verify loop, and reactive listeners.
Status: alpha / v0.1.1 published — core smoke passed against Ableton Live 12.4.1 on macOS and package validation is green. npm, GitHub Release .mcpb, and the MCP Registry are live. Glama has a listing, but its hosted release/deploy must be published separately from the Glama admin build flow; Smithery metadata is present and may lag indexing. API unstable. Don't use in production yet.>
Phase 8 status: slice 1 delivers read-only Max for Live/plug-in introspection and Link/remote status discovery. Deeper M4L control, VST3 sidecars, remote DAW integration and mobile companion work remain pending.
📚 Full documentation: pantani.github.io/ableton-mind/
Architecture (3 layers)
Claude/Cursor ──MCP/stdio──▶ ableton-mind (TS, Node 20+) ──TCP NDJSON JSON-RPC──▶ Remote Script (Python, inside Live)- `src/` — TypeScript MCP server. Tools, resources, prompts, TCP client, recipe runner, knowledge loader.
- `live/AbletonMind/` — Python Remote Script. TCP server on port
9876, dispatches JSON-RPC to LiveAPI. - `recipes/`, `src/knowledge/` — embedded JSON (drum kits, basslines, racks, device schemas).
Full spec in PLAN.md. Frozen contracts in _workspace/contracts/.
Highlights vs. existing MCP/OSC servers
| Capability | ahujasid/ableton-mcp | AbletonOSC + MCP wrapper | ableton-mind |
|---|---|---|---|
| MCP tools | 22 | ~30 | 36 |
| LOM coverage | ~10% | ~95% | ~100% |
| Knowledge base | none | none | 55 devices, scales, drum kits |
| Recipes | none | none | 14 across 7 categories |
| Verify loop | no | no | yes, integrated (`session_snapshot/diff`) |
| Render preview | no | no | yes (snapshot now, bounce planned) |
| Reactive listeners → MCP notifications | no | partial (OSC) | yes (7 events live) |
| Transactions (undo unitary) | no | no | yes |
| Automation envelopes | no | partial | complete (linear / hold) |
| Push 1/2/3 control | no | no | yes (pad/button/mode LEDs) |
| Docker | no | no | yes |
.mcpb 1-click | no | no | yes |
| Doctor CLI | no | no | yes |
Requirements
- Node 20+
- Ableton Live 12 (priority; Live 11 supported)
- macOS (primary), Windows (Phase 1 final)
Setup (source install)
npm install
npm run typecheck
npm run lint
npm run test
npm run buildInstall Remote Script (Python bridge)
Dev mode (symlink):
node scripts/install-remote-script.mjs # creates symlink
node scripts/install-remote-script.mjs --check # status only
node scripts/install-remote-script.mjs --copy # full copy (CI / snapshot)Manual:
- macOS: copy
live/AbletonMind/to~/Music/Ableton/User Library/Remote Scripts/AbletonMind/ - Windows: copy to
~/Documents/Ableton/User Library/Remote Scripts/AbletonMind/
Then Live → Preferences → Link/Tempo/MIDI → Control Surface → AbletonMind.
Smoke test: docs/smoke-test.md.
Run the MCP server
npm run build
node dist/index.jsEnv vars:
| Var | Default | ||||
|---|---|---|---|---|---|
ABLETON_MIND_HOST | 127.0.0.1 | Python bridge host | |||
ABLETON_MIND_PORT | 9876 | Bridge TCP port | |||
ABLETON_MIND_TIMEOUT_MS | 5000 | Per-request timeout | |||
ABLETON_MIND_MAX_FRAME_BYTES | 1048576 | Max incoming JSON-RPC frame | |||
ABLETON_MIND_MAX_PENDING_REQUESTS | 128 | Max in-flight JSON-RPC calls | |||
ABLETON_MIND_LOG_LEVEL | info | debug \ | info \ | warn \ | error |
Local copilot
Run a local LLM against a curated subset of the same Ableton tools:
ollama pull qwen2.5:3b # optional; the UI can pull too
node dist/index.js chat # opens http://127.0.0.1:4142
node dist/index.js ask "What is in this set?"The default tier is read-only (safe). Use --write for simple changes or --creative for recipes/browser load.
| Var | Default | |||
|---|---|---|---|---|
ABLETON_MIND_LLM_BASE_URL | http://127.0.0.1:11434/v1 | OpenAI-compatible endpoint | ||
ABLETON_MIND_LLM_MODEL | qwen2.5:3b | Local model id | ||
ABLETON_MIND_LLM_TIER | safe | safe \ | standard \ | creative |
ABLETON_MIND_CHAT_PORT | 4142 | Browser UI port |
See Local copilot.
Doctor CLI
npx ableton-mind-doctorChecks Node version, Remote Script install, bridge port, knowledge base integrity, recipes.
Distribution
- npm:
npm install -g ableton-mind. - Claude Code plugin marketplace:
claude plugin marketplace add Pantani/ableton-mind, thenclaude plugin install ableton-mind@ableton-mind. - Claude Desktop one-click: download
ableton-mind-0.1.1.mcpbfrom the v0.1.1 GitHub Release. - MCP Registry:
io.github.Pantani/ableton-mindis active in the official registry. - Glama: listed at glama.ai/mcp/servers/Pantani/ableton-mind; hosted release is separate from the GitHub Release and still requires the Glama admin deploy + Make Release flow.
- Source:
npm ci && npm run build && npm run install:remote-script. - Docker:
docker build -t ableton-mind . && docker run --rm -i --network host ableton-mind. - Smithery:
smithery.yamlis ready for listing/indexing.
Roadmap
See PLAN.md §12 and _workspace/PROGRESS.md.
| Phase | Status |
|---|---|
| 0 — Spike | ✅ real smoke pass |
| 1 — ahujasid parity | ✅ 22/22 |
| 2 — Listeners | ✅ 7 events |
| 3 — Knowledge | ✅ 55 devices |
| 4 — Automation envelopes | ✅ |
| 5 — Preview/verify | ✅ snapshot+diff (bounce planned) |
| 6 — Push | ✅ pad/button/mode LEDs |
| 7 — Distribution | ✅ DXT/Docker/Smithery/CI/release ready |
| 8 — Long tail | 🔵 slice 1 delivered: read-only M4L/plug-in introspection + Link/remote status discovery; deeper M4L/VST3/remote DAW/mobile work pending |
License
MIT.