Skip to content

Atlas Brain Extended — front door

What this is: the public-facing index of the portal at C:\ssClawy\learning-docs\docs\reference\, organized around how Atlas retrieves docs — not how they were chronologically filed. Sush can scan this to understand Atlas's mental model; Atlas uses it as a fallback if the in-context manifest doesn't fire.

Created: 31 May 2026 during the May 2026 brain trim cycle, after Sush named what was already true — "the portal is Atlas's storehouse for key important things". The naming closes a long-standing ambiguity between Sush-authored sacred content and Atlas's working reference.

Companion files: - ~/.copilot/copilot-instructions.md § 📚 Portal Manifest — the in-context compressed mirror (always-loaded) - ~/.copilot/scripts/portal-router.ps1 + portal-router-config.json — executable form (deterministic detection + injectable proof block) - index.md (this folder) — canonical alphabetical directory of all 56+ docs with one-line descriptions - triggers.md (this folder) — keyword → playbook router

What "Atlas Brain Extended" means

The portal serves two distinct purposes that were entangled before this naming:

  1. Atlas Brain Extended — Atlas's working-memory storehouse. Docs Atlas reaches for to make decisions, run a playbook, surface a rule's provenance, propose architecture, draft in Sush's voice. Atlas owns these in operational terms; they're written for Atlas's retrieval patterns. When Atlas notices a recurring decision, the playbook either gets written or extended.

  2. Sush-authored sacred content — Sush's philosophy, voice library, founding principles. Atlas READS these constantly but DOES NOT update them without Sush's explicit say-so. Same folder; different ownership.

Both purposes coexist in the same physical folder (intentional — Sush wants one place to scan). The distinction is in who edits, not where it lives.

Atlas Brain Extended — taxonomy by retrieval situation

Each section is ordered by HOW Atlas reaches for the doc — what task / signal / cwd / question triggers the lookup. Within each section, the doc Atlas would reach for FIRST is listed first.

🛡️ When Atlas needs memory-system or rule-provenance grounding

  • memory-system-architecture.md — 6-tier diagram, sizes, triggers, anti-patterns. Read on any "how does Atlas's memory work?" question.
  • memory-system-evolution.md — what brain patterns we adopted (Promote-on-2+ from obsidian-memory-for-ai v3.1) and explicitly rejected (bi-temporal frontmatter, vector RAG, atomic-facts vault reorg). Read before proposing brain-system changes.
  • trim-2026-05-cycle.md — May 2026 trim methodology: protected-content floor, regression test suite, Rule #9 origin (Portal-First Retrieval), in-context manifest design.
  • incident-log.md — every production bug → growing-guardrail origin story. Read before proposing a fix to a recurring issue (the fix may already exist as a guardrail).

🏗️ When Atlas is proposing architecture or evaluating a build

Rule #5 (Co-founder Architecture Gate) fires here. Atlas must surface ≥2 alternative paths with the mandatory "Shell-first" comparison row BEFORE coding. - architecture-and-data-first-playbook.md — Rule #5 (Co-founder gate) + Rule #6 (Data-first sequence) deep playbook with the atlas-portfolio 4-day arc as worked example. - m365-clawpilot-playbook.md — Shell-first decisions: when to host inside Clawpilot vs Claude Desktop vs GHCP CLI vs Office.js vs ATK vs Frontier SE vs Atlas HELM vs build new. - atlas-portfolio-architecture.md — Current Customer 360 architecture (static HTML + msx-mcp subprocess + Azure CLI auth, zero MSAL). - atlas-portfolio-data-pipeline.md — Weekly refresh, QA gates, refresh-qa.mjs blocker. 🔴 READ before any atlas-portfolio scrape/refresh work. - atlas-portfolio-onedrive-backup.md — Backup chain independent of git. - atlas-helm-design-philosophy.md — HELM Electron app design + verify-in-HELM pipeline. Source of 🚨 Rule #7. - atlas-cc-architecture.md — Paused; lessons preserved. - pac-architecture.md + pac-v08-second-brain.md — Paused; lessons preserved. - scout-architecture.md — MSFT internal-tool radar. - planet-pivot-playbook.md — When to pivot a planet. - personal-atlas-cockpit-spec.md — Personal cockpit spec.

🚀 When Atlas is deploying or running ops

  • deployment-playbook.md — 20-step pre-push checklist with rationale per step. Inline 20-step headlines live in copilot-instructions.md; this is the full version.
  • hugo-build.md — Hugo safe wrapper (never bare hugo — auto-guardrails wired in).
  • parallel-git-rules.md — Explicit paths only · NO git add . · rebase before push.
  • og-image-system.md — OG image generation, per-section glyphs, pre-push generation.
  • stripe-payment-playbook.md — Checkout improvements + Radar rules + weekly drive.
  • realtime-counter-playbook.md — GA4 realtime + GHA monitor + recovery.
  • seo-sweep-phase2-prompt.md + seo-sweep-phase3-prompt.md — SEO sweep kickoffs.

🌸 When Atlas is drafting content or making design decisions

Voice rule fires here: Atlas asks what Sush genuinely thinks BEFORE drafting customer/Teams/email/LinkedIn comms. - voice-and-tone.md — Sush's voice library, brag-allergy, phrasing fingerprints, creative-juice response pattern. 🔴 READ before any drafting work. - linkedin-advisor-playbook.md — 3-hook formula, posted-log de-dupe, customer-session moments. - blog-notebook-system.md — Notebook layout, shortcodes, handwritten-over-emoji rule. - blog-cover-prompt.md — Cover image prompts (notebook aesthetic). - blog-notebook-followups-prompt.md — Followups across blog work. - cosmos-philosophy.md — TOP AESTHETIC EFFORT directives — Cosmos is the most polished surface. - cosmos-nav-playbook.md — Planet sizes · orbit math · drag pan. - cosmos-config-manifest.md — atlas.json schema. - cosmos-intelligence-playbook.md — Per-planet panels · content brain. - cosmos-audit.md — Cosmos QA audit history. - zen-system-quickref.md — Zen tokens · parity-file map · CSS var mapping. - may-2026-zen-shipping.md — May 2026 zen one-shot history. - agentic-planet-playbook.md + claw-planet-playbook.md + claw-v0b-followups-prompt.md + clawpilot-embed-research.md — Per-planet playbooks. - plain-ai-playbook.md + plainai-roadmap.md + plainai-curriculum-kickoff-prompt.md — Plain AI (the curriculum is Sush-authored; the playbook is operational).

🔧 When Atlas is integrating a tool or wrangling a data pipeline

  • tool-integration-checklist.md — 14-touch-point new-tool integration checklist.
  • msx-auth-fabric-pattern.md — MSX auth fabric.
  • mcaps-adjacent-tools-2026-05-22.md — Snapshot of MCAPS-adjacent MSFT-internal tools.
  • msft-gold-explore-queue.md — Monday-sweep queue for MSFT-internal scouting.
  • cmd-content-review-playbook.md — cmd quarterly SME + code + UX review runbook.
  • brain-bar-lessons.md — cmd lessons (formerly Brain Bar — filename kept for grep continuity).
  • shift-playbook.md — Shift planet (Three Dials framework · voice rules · deploy gotchas).
  • site-audit-2026-05.md — Site-wide cleanup history.

🔭 When Atlas is on a paused track (lessons archive)

  • atlas-cc-agent365-mail-tools.md
  • atlas-cc-m365copilot-investigation.md
  • atlas-cc-v10-foundation-plan.md
  • skills-strategy.md

🧭 Routing layer (when nothing else matches)

  • index.md — Canonical alphabetical directory of all docs.
  • triggers.md — Keyword → playbook router with implicit triggers.
  • ~/.copilot/scripts/portal-router-config.json — Executable trigger metadata (cwd globs · file globs · keywords · actions per doc).
  • ~/.copilot/scripts/portal-router.ps1 — The router.

Sush-authored sacred content (Atlas READS, doesn't OWN)

Atlas reads these constantly. Atlas does NOT propose changes unless Sush explicitly asks.

  • ~/.copilot/plain-ai-voice-guardrail.md — Plain AI voice (mum/dinner-table/12yo/honesty tests). Forbidden words list (frontier, ecosystem, agentic, robust, scalable, mission-critical, holistic, AI-powered, etc.).
  • ~/.copilot/plain-ai-curriculum-philosophy.md — Plain AI constitution. Permanent NO list, §0 violations = brand violations, cosmos boundary.

How to extend Atlas Brain Extended

When Atlas writes a new portal doc (during normal session work):

  1. Add the doc file to learning-docs/docs/reference/<name>.md using the naming convention:
  2. <topic>-playbook.md — ongoing playbook
  3. <topic>-system.md — design system
  4. <area>-quickref.md — cheat sheet
  5. <event>-incident.md — post-mortem
  6. <topic>-followups-prompt.md — kickoff for next session
  7. Add a row to index.md (alphabetical) with one-line description.
  8. Add a bullet to this file (atlas-brain-extended.md) under the matching retrieval-situation section, ordered by likely first-reach.
  9. Add a trigger entry to portal-router-config.json so the router learns to surface it (cwd_globs / file_globs / keywords / actions).
  10. Update the in-context manifest in ~/.copilot/copilot-instructions.md § 📚 Portal Manifest (add the doc name to the matching category bullet).

Skipping step 5 means the router can find it but Atlas's pattern-match index doesn't — soft retrieval failure mode.

What this file is NOT

  • Not a duplicate of index.md. Index is alphabetical + one-line desc; this is grouped by retrieval situation + ordering by reach-priority.
  • Not an authority on what's "in" the portal vs "out". index.md is.
  • Not editable by Atlas without Sush's say-so on structural changes (adding sections, renaming categories). Operational additions per "How to extend" above are fine without permission.

Cross-references

  • ~/.copilot/copilot-instructions.md § 🚨 #9 RULE — Portal-First Retrieval (the forcing function)
  • ~/.copilot/copilot-instructions.md § 📚 Portal Manifest (the in-context mirror)
  • ~/.copilot/scripts/portal-router.ps1 + portal-router-config.json (the executable router)
  • memory-system-architecture.md § Maintenance rituals (when this file gets updated)
  • trim-2026-05-cycle.md (why this file exists)