SecondBrain/cc-os-hub.md

---
type: hub
title: cc-os — personal cross-project memory system for Claude Code
summary: Navigation hub for cc-os, the personal cross-project memory system for Claude Code (episodic memsearch + semantic SecondBrain vault via Graphify) and its family of global plugins (os-vault, os-adr, os-orchestration, os-doc-hygiene) — design, conventions, and eval results.
tags:
  - type/hub
  - project/cc-os
  - domain/knowledge-graphs
  - domain/llm-evaluation
  - tool/graphify
  - tool/claude-code
scope: global
last_updated: 2026-07-06
date: 2026-07-06
source: cc-os
---

# cc-os

`cc-os` (`/home/jared/dev/cc-os`) is a design-and-implementation repo for a personal,
cross-project memory system for Claude Code, built for a multi-client freelancer. It splits
memory into two systems — **episodic** (memsearch, "what happened, when") and
**semantic/knowledge** (this vault, "how do we…", queried via Graphify) — and ships a family of
global Claude Code plugins (`os-vault`, `os-adr`, `os-orchestration`, `os-doc-hygiene`) that
operationalize conventions across every project, not just cc-os itself.

## Design & decisions

- [[2026-06-09-claude-code-memory-system-design]] — the two-layer architecture (episodic +
  semantic), hook-injected retrieval, freshness model.
- [[2026-06-09-claude-code-memory-system-adrs]] — the 14-decision ADR log behind that design.
- [[2026-06-09-graphify-evaluation-as-knowledge-layer]] — why Graphify was adopted as the
  knowledge-layer engine.
- [[vault-conventions]] — the vault's frontmatter contract, note types, and tag taxonomy (single
  source of truth; don't duplicate it elsewhere).
- [[cc-os-plugin-skill-naming-convention]] — `os-[domain]` plugin naming, verb-first skills, and
  the `name:`-field-in-SKILL.md registration gotcha.
- [[vault-backlog-pilot-plan]] / [[backlog-system-options-research]] — the cross-project backlog
  decision (Planka CE + thin Ruby CLI) adjacent to the vault's knowledge-layer role.

## Graphify handbook

The verified reference handbook behind cc-os's chosen knowledge-graph engine:
[[2026-06-08-graphify-conceptual-model]] (mental model) ·
[[2026-06-08-graphify-installation-setup]] ·
[[2026-06-08-graphify-overview-concepts]] ·
[[2026-06-08-graphify-ingesting-code-ast]] ·
[[2026-06-08-graphify-ingesting-docs-knowledge]] ·
[[2026-06-08-graphify-extraction-model-options]] ·
[[2026-06-08-graphify-local-models-backends]] ·
[[2026-06-08-graphify-querying-god-nodes]] ·
[[2026-06-08-graphify-token-economics-updates]] ·
[[2026-06-08-graphify-workflows-use-cases]] ·
[[2026-06-08-graphify-best-practices-checklist]].

## Eval methodology and results

cc-os plugins are validated by held-out, unprompted-behavior evals rather than prompted demos.
Start at [[eval-methodology-hub]] for the general methodology (ladder design, wording-loop
procedure, production feedback loop). Concrete results:

- **os-adr:** [[2026-07-03-os-adr-eval-b-grid-results-and-observations]] (2026-07-03 baseline;
  note: a duplicate slug-only file `os-adr-eval-b-grid-results-and-observations.md` also exists
  and is empty — likely a stale rename artifact, flagged here but not fixed as out of scope) →
  [[os-adr-eval-b-wording-experiment-hypotheses]] (wording loop, sonnet 8/8 / haiku 7/8) →
  [[os-adr-eval-c-frozen-grid-results]] (frozen ambiguity-ladder grid, 12/12 cells pass).
- **os-orchestration:** [[os-orchestration-ws1-session-audit-results]] (10-session IRL audit;
  found the `CLAUDE_CODE_SUBAGENT_MODEL` env override) → [[os-orchestration-eval-baseline-grid-results]]
  (E1–E3 baseline grid).
- **os-vault:** [[os-vault-write-eval-baseline-grid-results]] (2026-07-06 untuned baseline:
  positives 1/19, negatives 18/18; L1 knowledge routes to Claude Code auto-memory instead of
  the vault — routing plus L2/L3 triggering are the wording-loop targets; harness at
  `cc-os/plugins/os-vault/eval/`, wording loop not yet started).

## Related

- [[eval-methodology-hub]] — general evaluation methodology (not cc-os-specific, but built on
  and by cc-os plugin work).