brain-snapshot
Skill7 starsChanged 19 days agoMITView on GitHub
Refresh the runtime context payload other skills read at task time. Say "refresh the brain snapshot", "rebuild context", or "update the snapshot". Also fires when wiring a new skill to the snapshot or when checking which fields skills can rely on. The snapshot is generated by `scripts/brain-snapshot.py` and lives at `brain/.snapshot.md` (gitignored, per-user).
In this repository100 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- notion-package/CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- add-handsSkill
- add-mouthSkill
- add-voiceSkill
- approval-gatesSkill
- auditSkill
- backupSkill
- blind-spot-reviewSkill
- bottleneck-diagnosticSkill
- brain-logSkill
- brain-passSkill
- brand-interviewSkill
- brand-voice-interviewSkill
- business-context-loaderSkill
- campaign-from-themeSkill
- capture-sweepSkill
- catch-upSkill
- client-updateSkill
- connectSkill
- content-repurposerSkill
- content-startSkill
- context-persistenceSkill
- cross-linkSkill
- data-securitySkill
- decision-frameworkSkill
- decisions-startSkill
- delivery-startSkill
- email-drafterSkill
- employee-reviewSkill
- finance-importSkill
- forcing-questionsSkill
- founder-coachingSkill
- founder-next-moveSkill
- founder-os-setupSkill
- founder-reviewSkill
- founder-scope-challengeSkill
- github-opsSkill
- handoff-protocolSkill
- hireSkill
- historySkill
- housekeepingSkill
- ingestSkill
- knowledge-captureSkill
- legal-complianceSkill
- linkedin-brand-directionSkill
- linkedin-network-scanSkill
- linkedin-postSkill
- linkedin-power-auditSkill
- linkedin-startSkill
- linkedin-warm-revivalSkill
- lintSkill
- list-prunerSkill
- log-archiveSkill
- log-replySkill
- meeting-prepSkill
- memory-passSkill
- menuSkill
- The first 60 shown.
The file
Files
Brain Snapshot
Runs on: local-writes - creates or edits files in your OS folder; needs an agent with write access.
The brain snapshot is a small markdown payload (~200 tokens) that captures what is true right now: voice rules, brand fields, open flags, this week's commitments, recent decisions, and staleness state. Skills read it at task time so their output reflects the latest state of the OS instead of starting cold.
Where it lives
- Generator:
scripts/brain-snapshot.py(Pure Python stdlib, no runtime deps). - Output:
brain/.snapshot.md(gitignored, per-user state). - Mirror copy for installs:
templates/scripts/brain-snapshot.py.
Generate or refresh the snapshot
python scripts/brain-snapshot.py --writeWrites the snapshot to brain/.snapshot.md. Re-run any time the underlying files change: usually after a /dream digest, after rolling the daily anchor, or at the start of a long session.
To inspect without writing:
python scripts/brain-snapshot.pySnapshot format
Deterministic field order. Same inputs produce identical output (modulo the date: line). Sections always render: missing source files emit [unavailable] rather than crashing.
# Brain snapshot
date: YYYY-MM-DD
staleness:
daily_anchor: fresh | stale (N days past) | [unavailable]
weekly: fresh | stale (N days past) | [unavailable]
weekly-must-do: fresh | [unavailable - weekly file stale] | [unavailable - weekly file missing]
## Voice (key rules)
- rhythm: <value or [NOT SET]>
- opening_style: <value or [NOT SET]>
- closing_style: <value or [NOT SET]>
- contractions: <value or [NOT SET]>
- reading_level: <value or [NOT SET]>
- preferred: <comma-joined list or [NOT SET]>
- banned: <comma-joined list or [NOT SET]>
## Brand (key fields)
- display_name: <value or [NOT SET]>
- primary_color: <hex or [NOT SET]>
- primary_font: <font or [NOT SET]>
## Open flags (top 3)
- <flag header 1>
- <flag header 2>
- <flag header 3>
## This week (must do)
- <item 1>
- <item 2>
- <item 3>
## Recent decisions (last 3)
- <decision header 1>
- <decision header 2>
- <decision header 3>How skills consume the snapshot
A skill that opts in reads brain/.snapshot.md before producing output. The opt-in pattern is consistent across all consumers:
1. Read brain/.snapshot.md if it exists.
2. If missing, OR its date: line is more than 3 days old, run:
python scripts/brain-snapshot.py --write
Then read it. A stale cache read as current aims output at last month's state.
3. If the snapshot script is also missing (older install), proceed using only profile files.
Do not block.
4. Apply the snapshot as runtime context. Do not surface every field in every output.
Use judgment about which fields belong in which output.What the snapshot is NOT
- It is not a database. It is a flat markdown payload regenerated from source files on demand.
- It is not authoritative. The source files (flags.md, weekly-commitments.md, decisions.md, voice-profile.yml, brand-profile.yml) remain the single source of truth.
- It is not a replacement for
query.py. Query is for retrieval across the full corpus. Snapshot is for "what is true right now." - It is not auto-fired. Skills consume it on opt-in. Setup does not require generating it.
Failure modes and recovery
- Script missing (older install). Skills fall back to profile-only context. No output is blocked.
- Source file missing. That section emits
[unavailable]and others render normally. - Source file unparsed (e.g. weekly file lacks the
## Week of YYYY-MM-DDheader). The relevant staleness field shows[unparsed - ...]. The OS still functions; the user can fix the header at their convenience. - Snapshot stale. Re-run with
--write. The snapshot is cheap to regenerate.
Privacy
brain/.snapshot.md may contain personal fields (display_name, voice rules, decision headers). It is gitignored by default. If you sync the OS folder across machines, the snapshot regenerates locally on each machine.
Determinism
For test stability, the script accepts --today YYYY-MM-DD to override the current date. This is test-only. Production runs use date.today().
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

