How do I install the Od Memory skill?

Run npx skillmds@latest add agsqaintern/od-memory 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 Od Memory 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 Od Memory?

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 Od Memory free to use?

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

Who published Od Memory?

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

Od Memory by agsqaintern — Agent Skill

Name: Od Memory
Author: AgsQaintern

Od Memory

Operate Open Design daemon memory — the durable markdown fact store under the daemon data root that is injected into future agent prompts. Use when the user mentions od memory, daemon memory, memory tree, MEMORY.md, user profile, memory rules, memory extraction, or wants to list/edit/move memory entries, set profile, add rules, or toggle memory config via CLI or /api/memory.

AgsQaintern Updated Sep 22, 2026

File contents

Open Design daemon memory

Durable product memory owned by the OD daemon — not Cursor chat context.

What it is

Filesystem store under <OD_DATA_DIR>/memory/:
- MEMORY.md — index (one bullet per fact)
- <type>_<slug>.md — entry body + frontmatter
- .config.json — feature switches
Accepted entries are injected into future daemon / BYOK agent prompts.
Surfaces: Settings UI, /api/memory/*, od memory …

Types (buckets)

Type	Role
`profile`	Singleton `user_profile` — who / how; PRE-loop brief expansion
`user`	Stable user prefs / identity facts
`feedback`	Corrections and taste notes
`project`	Project-specific facts
`reference`	External refs / docs pointers
`rule`	Assertion + check — POST-loop self-verify rubric

Sources (frontmatter source:): manual · heuristic · llm · connector · brand · annotation

Prefer CLI over scraping UI

Daemon must be running. Use --json when scripting. Optional --daemon-url <url>.

od memory tree list --json
od memory tree view <id> --json
od memory tree edit <id> --name "…" --description "…" --type user|feedback|project|reference --body "…"
od memory tree move <id> --type user|feedback|project|reference

od memory profile show --json
od memory profile set --field "Label=Value" [--prompt-file path|-]

od memory rule list --json
od memory rule add --name "…" --assertion "…" --check "…"
od memory rule suggest --note "…"   # proposals only; keep with rule add

od memory verify list --json
od memory verify clear --json

od memory config --json
od memory config --enabled true|false --extraction true|false --profile true|false --rewrite true|false --verify true|false

--extraction maps to chat auto-extraction. Master --enabled false turns the whole feature off.

HTTP (same contract)

GET/PATCH /api/memory/tree (+ …/tree/:id)
Entry CRUD under /api/memory/…
Config under /api/memory/config
SSE: /api/memory/events

Do not invent a second store. Do not write under cwd-relative legacy paths — memory lives under the resolved daemon data root (OD_DATA_DIR → RUNTIME_DATA_DIR).

Agent behavior

Inspect before edit — tree list / view so you don't duplicate facts.
Prefer one clear entry over many overlapping bullets.
Profile for stable identity/workflow; feedback for corrections; rule only when there is a checkable assertion.
Confirm before destructive config toggles or verify clear.
Do not confuse this with Cursor Memories / chat history — those are IDE-side and do not feed OD prompts.

When coding memory itself

Source of truth: apps/daemon/src/memory.ts, routes in apps/daemon/src/routes/memory.ts, contracts in packages/contracts/src/api/memory.ts. CLI help in apps/daemon/src/cli.ts (od memory). Keep UI + CLI + HTTP parity.

---name: od-memorydescription: Operate Open Design daemon memory — the durable markdown fact store under the daemon data root that is injected into future agent prompts. Use when the user mentions od memory, daemon memory, memory tree, MEMORY.md, user profile, memory rules, memory extraction, or wants to list/edit/move memory entries, set profile, add rules, or toggle memory config via CLI or /api/memory.---# Open Design daemon memoryDurable product memory owned by the OD daemon — **not** Cursor chat context.## What it is- Filesystem store under `<OD_DATA_DIR>/memory/`: - `MEMORY.md` — index (one bullet per fact) - `<type>_<slug>.md` — entry body + frontmatter - `.config.json` — feature switches- Accepted entries are injected into **future** daemon / BYOK agent prompts.- Surfaces: Settings UI, `/api/memory/*`, `od memory …`## Types (buckets)| Type | Role ||------|------|| `profile` | Singleton `user_profile` — who / how; PRE-loop brief expansion || `user` | Stable user prefs / identity facts || `feedback` | Corrections and taste notes || `project` | Project-specific facts || `reference` | External refs / docs pointers || `rule` | Assertion + check — POST-loop self-verify rubric |Sources (frontmatter `source:`): `manual` · `heuristic` · `llm` · `connector` · `brand` · `annotation`## Prefer CLI over scraping UIDaemon must be running. Use `--json` when scripting. Optional `--daemon-url <url>`.```bashod memory tree list --jsonod memory tree view <id> --jsonod memory tree edit <id> --name "…" --description "…" --type user|feedback|project|reference --body "…"od memory tree move <id> --type user|feedback|project|referenceod memory profile show --jsonod memory profile set --field "Label=Value" [--prompt-file path|-]od memory rule list --jsonod memory rule add --name "…" --assertion "…" --check "…"od memory rule suggest --note "…" # proposals only; keep with rule addod memory verify list --jsonod memory verify clear --jsonod memory config --jsonod memory config --enabled true|false --extraction true|false --profile true|false --rewrite true|false --verify true|false````--extraction` maps to chat auto-extraction. Master `--enabled false` turns the whole feature off.## HTTP (same contract)- `GET/PATCH /api/memory/tree` (+ `…/tree/:id`)- Entry CRUD under `/api/memory/…`- Config under `/api/memory/config`- SSE: `/api/memory/events`Do not invent a second store. Do not write under cwd-relative legacy paths — memory lives under the resolved daemon data root (`OD_DATA_DIR` → `RUNTIME_DATA_DIR`).## Agent behavior1. **Inspect before edit** — `tree list` / `view` so you don't duplicate facts.2. **Prefer one clear entry** over many overlapping bullets.3. **Profile** for stable identity/workflow; **feedback** for corrections; **rule** only when there is a checkable assertion.4. **Confirm** before destructive config toggles or `verify clear`.5. Do not confuse this with Cursor Memories / chat history — those are IDE-side and do not feed OD prompts.## When coding memory itselfSource of truth: `apps/daemon/src/memory.ts`, routes in `apps/daemon/src/routes/memory.ts`, contracts in `packages/contracts/src/api/memory.ts`. CLI help in `apps/daemon/src/cli.ts` (`od memory`). Keep UI + CLI + HTTP parity.

Frequently asked questions