← Vault Index
Source: frameworks/kit-new-client/02-new-client-kit-terminology.md

02 — TERMINOLOGY: New Client Kit

Locked vocabulary for the New Client Kit. These terms have specific meanings in this system. This file wins if any other file uses different terminology.


Workspace Terms

TermDefinitionNOT This
Client workspaceThe entire repo for a single client. Contains all client-specific files, identity files, and directory structure. One workspace = one repo = one client.Not "project" (a project is a subdirectory within the workspace)
CLAUDE.mdThe identity file at the repo root. Tells Claude who the client is, what the engagement looks like, where files are, what frameworks to use, and what rules apply. The single source of truth for how Claude operates in this workspace.Not "config" or "settings" (those are .claude/settings.local.json)
WORKFLOW.mdThe processing instructions file at the repo root. Documents the weekly rhythm, pre-processing requirements, and reference material for the 9-step post-session cascade. The cascade skills in .claude/skills/ drive execution; WORKFLOW.md is the human-readable documentation.Not "playbook" or "runbook"
Reference dataThe canonical source of truth for all proper nouns — names, companies, tools, team members, transcript corrections. Lives in reference-data/. Overrides every other source.Not "context" or "background"
Sessions folderPermanent record of session transcripts, JSON analysis files, and recaps. Files land here and never move. One transcript + one JSON per session.Not "inputs" (inputs is for non-session material)
Inputs folderStaging area for external non-session material — client docs, status updates, constraint briefs. Files move to archive/ once processed (extracted and incorporated into a deliverable).Not "sessions" or "uploads"
Drafts folderActive production staging — email drafts, agendas, in-progress work. Files move to ready/ when finished or archive/ if abandoned.Not "outputs" or "builds"
Ready folderFinished work waiting to be sent or published. Files move to archive/ when delivered.Not "outputs"
Archive folderFinal state for all retired, consumed, or abandoned work. Nothing leaves archive.Not "trash" — these are records
Reference coreFour identity files (soul.md, offer.md, audience.md, voice.md) in reference/core/. Client portrait — who they are, what they offer, who they serve, how they sound. Separate from reference data.Not "reference data" (reference data is proper nouns/tools/names)

Brand & Customization Terms

TermDefinitionNOT This
Brand kitThree files that define a client's visual and verbal identity: HTML reference (brand guide), brand skill (production rules), brand QC (quality gate). All three live in brand/.Not a single file — always 3 files
Client-specific skillA markdown file in skills/ that extends a vault framework with client-specific rules. It references the vault framework — it does not duplicate it.Not a copy of the vault skill
Client-specific QCA markdown file in qc-agents/ that layers client-specific quality checks on top of vault QC. Additive, not replacement.Not a standalone QC — always layers on vault QC

Workspace Maturity Terms

TermDefinition
Minimum viable workspaceA workspace that works using vault frameworks only. Has the full directory structure, CLAUDE.md, WORKFLOW.md, settings, reference data — but empty brand/, skills/, qc-agents/ directories. JB (Jenn Barr) is the live example.
Fully mature workspaceA workspace with brand kit, client-specific skills, and client-specific QC agents. RC (Ruben Cruz) is the live example. All three categories populated.

File Naming Terms

TermDefinitionExample
Preferred nameThe client's first name, used as the prefix for all client-specific files. Lowercase in file names.ruben, jenn
Client slug2-3 lowercase character code derived from client initials. Used only in the repo name.rc, jb
Repo nameAlways follows aos-client-XX pattern where XX is the client slug.aos-client-rc, aos-client-jb

Workflow Terms

TermDefinition
Post-session cascadeThe 9-step auto-advancing skill chain in .claude/skills/ that processes session inputs into all living documents. Steps: recap-email → recap-internal → recap-json → update-cpm → update-masterplan → update-projectplan → update-blueprint → draft-actions → session-agenda. Each step auto-advances to the next. Human review gate between every step. Installed during Mode 1 (Day 0 scaffold).
Cascade skillA skill file in .claude/skills/[step-name]/SKILL.md that handles one step of the post-session cascade. Contains: what to read, what to produce, where to push, auto-advance instruction. At Day 0, references vault frameworks for processing logic. After Mode 3, may reference client-specific production skills instead.
Human review gateA mandatory stop point between cascade steps. Claude produces output, then stops. Kathryn reviews and approves before the next step runs. Every step has one. No exceptions.
Session openerThe /start skill in .claude/skills/start/. Vault-inbox check, file move review, workflow position check, and kanban state confirmation before any production work begins.
Kanban flowThe file lifecycle: sessions/ (permanent) · inputs/archive/ · drafts/ready/archive/. Every file has a stage. Files move forward, never backward.
Weekly updateThe processing sequence triggered by session inputs. Runs the 9-step post-session cascade. Kathryn says /recap-email and the cascade auto-advances through all steps with human review gates.
Processing sessionA single Claude Code session where Kathryn processes a client's weekly inputs through the cascade. May process one or multiple session transcripts.

Mode Terms

TermDefinition
Day 0The day a client signs. Mode 1 runs on Day 0 to create the workspace.
Session 1The first advisory session with the client. Mode 2 runs after Session 1 to populate the workspace.
Maturity upgradeAny Mode 3 action — creating a brand kit, adding a skill, adding a QC agent. Each is independent and incremental.