How do I install the Openbrain Operator skill?

Run npx skillmds@latest add igorwarzocha/openbrain-operator in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Openbrain Operator safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Openbrain Operator?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Openbrain Operator free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Openbrain Operator?

igorwarzocha (@igorwarzocha) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Openbrain Operator by igorwarzocha — Agent Skill

Name: Openbrain Operator
Author: igorwarzocha

Openbrain Operator

Operate and troubleshoot OpenBrain memory through the `brain` CLI and OpenBrain HTTP API. Use when the user asks to capture thoughts, run semantic search, review recent/history, check stats/health, tune retrieval, or diagnose why memory read/write is failing.

igorwarzocha 931b38f 4 files · 6.7 KB Updated Sep 17, 2026

File contents

OpenBrain Operator

This is the primary operator workflow and is optimized for localhost installations. Use this file as the default source of truth. Do not use repo-wide docs.

LAN Switch Rule (Required)

references/lan-operations.md
references/troubleshooting.md

Run:

brain health

If brain is unavailable:

npm run brain -- health

If health JSON shows "mode": "remote":
- Switch to LAN workflow by loading references/lan-operations.md.
- For this run, treat localhost instructions in this file as deleted.
- Continue with LAN commands only.
If health JSON shows "mode": "local":
- Stay in this file and run localhost workflow below.
If both CONVEX_URL and OPENBRAIN_REMOTE_URL exist, remote wins (lan).

Localhost Workflow (Default)

1. Preflight

node -v
npm -v
brain health

If brain is not found:

npm run brain -- health

Health must return JSON containing:

"ok": true
"mode": "local"

2. Core Operations

Capture:

brain capture "Decided to delay launch for QA blockers" --tags planning,qa

Search:

brain search "What did I note about QA blockers?" --limit 8 --threshold 0.2
brain search "What did I note about QA blockers?" today

Recent:

brain recent --limit 20
brain recent today

Stats:

brain stats

Remove:

brain remove --recent 1
brain remove --content "Decided to delay launch for QA blockers"
brain remove --query "the note about delaying launch"

3. Remove Validation

Inspect candidates:

brain recent --limit 20

Remove one thought:

brain remove --recent 1

Verify the thought is gone:

brain recent --limit 20
brain search "<same content>" --limit 20 --threshold 0.0

4. Retrieval Tuning

If too strict:

brain search "<query>" --limit 20 --threshold 0.0

If too noisy:

brain search "<query>" --limit 8 --threshold 0.3

Localhost Troubleshooting (Fast Path)

brain: command not found:

npm run link:global
which brain

Health fails:
- Run npm run brain -- health.
- If still failing, load references/troubleshooting.md.
401 unauthorized:
- Align OPENBRAIN_API_KEY.
- See references/troubleshooting.md.

Execution Contract

Execute only commands for the active mode.
Verify each step with concrete evidence (ok, IDs, counts, or HTTP responses).
On first failure, switch to references/troubleshooting.md and run the matching fix path.
Re-run health after each fix before continuing.

Response Contract

Report active mode (localhost or lan).
Show exact commands executed.
Show key outputs:
- Health status.
- Saved/removed thought summary when applicable.
- Search results or empty result.
If unresolved, provide one next action with the exact command.

---name: openbrain-operatordescription: Operate and troubleshoot OpenBrain memory through the `brain` CLI and OpenBrain HTTP API. Use when the user asks to capture thoughts, run semantic search, review recent/history, check stats/health, tune retrieval, or diagnose why memory read/write is failing.---# OpenBrain OperatorThis is the primary operator workflow and is optimized for `localhost` installations.Use this file as the default source of truth. Do not use repo-wide docs.## LAN Switch Rule (Required)- `references/lan-operations.md`- `references/troubleshooting.md`1. Run:```bashbrain health```If `brain` is unavailable:```bashnpm run brain -- health```2. If health JSON shows `"mode": "remote"`: - Switch to LAN workflow by loading `references/lan-operations.md`. - For this run, treat localhost instructions in this file as deleted. - Continue with LAN commands only.3. If health JSON shows `"mode": "local"`: - Stay in this file and run localhost workflow below.4. If both `CONVEX_URL` and `OPENBRAIN_REMOTE_URL` exist, remote wins (`lan`).## Localhost Workflow (Default)### 1. Preflight```bashnode -vnpm -vbrain health```If `brain` is not found:```bashnpm run brain -- health```Health must return JSON containing:- `"ok": true`- `"mode": "local"`### 2. Core OperationsCapture:```bashbrain capture "Decided to delay launch for QA blockers" --tags planning,qa```Search:```bashbrain search "What did I note about QA blockers?" --limit 8 --threshold 0.2brain search "What did I note about QA blockers?" today```Recent:```bashbrain recent --limit 20brain recent today```Stats:```bashbrain stats```Remove:```bashbrain remove --recent 1brain remove --content "Decided to delay launch for QA blockers"brain remove --query "the note about delaying launch"```### 3. Remove Validation1. Inspect candidates:```bashbrain recent --limit 20```2. Remove one thought:```bashbrain remove --recent 1```3. Verify the thought is gone:```bashbrain recent --limit 20brain search "<same content>" --limit 20 --threshold 0.0```### 4. Retrieval TuningIf too strict:```bashbrain search "<query>" --limit 20 --threshold 0.0```If too noisy:```bashbrain search "<query>" --limit 8 --threshold 0.3```## Localhost Troubleshooting (Fast Path)- `brain: command not found`:```bashnpm run link:globalwhich brain```- Health fails: - Run `npm run brain -- health`. - If still failing, load `references/troubleshooting.md`.- `401 unauthorized`: - Align `OPENBRAIN_API_KEY`. - See `references/troubleshooting.md`.## Execution Contract1. Execute only commands for the active mode.2. Verify each step with concrete evidence (`ok`, IDs, counts, or HTTP responses).3. On first failure, switch to `references/troubleshooting.md` and run the matching fix path.4. Re-run health after each fix before continuing.## Response Contract1. Report active mode (`localhost` or `lan`).2. Show exact commands executed.3. Show key outputs: - Health status. - Saved/removed thought summary when applicable. - Search results or empty result.4. If unresolved, provide one next action with the exact command.

Frequently asked questions