zenbin-brain

name: zenbin-brain description: Opinionated ZenBin second-brain workflow for agent memory. Use when asked to remember, save, recall, search, update, seed, maintain, or audit an agent's ZenBin brain/second brain/private memory; when publishing durable agent memory, journals, decisions, project state, operating conventions, lessons learned, open loops, or private notes to ZenBin; or when reading private sign-to-read memory pages from a ZenBin wiki index.

ZenBin Brain

Use ZenBin as an agent second brain: durable signed memory pages, private by default, discoverable through a public metadata-only _wiki index.

Prime Directive

Agent brain content is private sign-to-read by default.

Use public pages only for intentionally public specs, blog posts, references, demos, and docs. For memory, journals, project state, decisions, open loops, operating conventions, relationship/context notes, and lessons learned, publish private:

node <zenbin-publisher>/scripts/publish.js \
  --slug <slug> \
  --markdown <file.md> \
  --recipient me \
  --sign-to-read \
  --update-index

recipientKeyId / --recipient alone is not privacy. It only routes/list pages for a recipient. Private read access requires auth.signToRead: true / --sign-to-read.

What Belongs in the Brain

Save durable, reusable context:

• decisions and rationale
• current project state
• open loops and next actions
• operating conventions and preferences
• lessons learned and mistakes to avoid
• relationship/context notes that help future work
• compact research summaries worth recalling later

Do not save raw credentials, private keys, tokens, passwords, or unnecessary sensitive dumps. Sanitize first.

Write Workflow

1. Decide whether the note is durable. If it is temporary or noisy, keep it local instead.
2. Create a concise Markdown note with a clear title, date, summary, and links/backlinks.
3. Publish with --recipient me --sign-to-read --update-index.
4. Verify:
   • unsigned GET to the page returns 401
   • signed GET returns 200
   • _wiki has the entry with data-visibility="private"
5. Mention the slug/URL and verification result.

Suggested local draft location: a configurable brain draft folder such as $ZENBIN_BRAIN_DIR/<slug>.md. In OpenClaw workspaces, ~/workspace/zenbin-memory/<slug>.md is a convenient convention.

Recall Workflow

1. Read the public index: https://<subdomain>.zenbin.org/_wiki.
2. Scan <section data-wiki-entry> entries by title, tags, description, category, and links.
3. For public entries, read the page normally.
4. For data-visibility="private", use a signed GET with the recipient key.
5. Synthesize the answer. Cite page slug/source when useful.

Helper for private reads:

node scripts/signed-get.js <slug> --subdomain <name> --format md
# or set ZENBIN_SUBDOMAIN=<name>

Index Rules

_wiki should stay public and contain metadata only for private brain pages:

<section data-wiki-entry
         data-id="project-state"
         data-tags="project,state,open-loops"
         data-category="log"
         data-visibility="private">
  <h3>Project State</h3>
  <p>Current private project state and next actions. Sign to read.</p>
</section>

Never copy private page content into _wiki. The index should reveal enough to find the right page, not the private content itself.

Maintenance Workflow

Periodically:

1. Read _wiki and inspect brain entries.
2. Merge duplicates and stale fragments.
3. Promote repeated daily notes into durable brain pages.
4. Update open-loop pages as work completes.
5. Re-run index update and verify private entries remain metadata-only.

Helper Script

scripts/signed-get.js supports:

• --subdomain <name> or ZENBIN_SUBDOMAIN for slug inputs
• --format html|md
• --json for structured output
• --verify to check unsigned 401 plus signed 200

Supporting Skills

• Use zenbin-publisher for publishing/API mechanics.
• Use zenbin-wiki for the wiki index convention.
• Use this skill for the opinionated memory policy and brain workflow.

For the repo-level design note, see docs/specs/zenbin-brain-skill.md.