Fix install symlinks (#5)

* Add ./create installer to symlink skills into ~/.claude/skills

A plain `git clone` into ~/.claude/skills/creativestack produces a
nested structure that Claude Code does not discover, since skills must
live at ~/.claude/skills/<name>/SKILL.md rather than one level deeper.

Add a `create` script that symlinks each skills/<name>/SKILL.md into
~/.claude/skills/<name>/SKILL.md and each agents/<name>.md into
~/.claude/agents/<name>.md, so all 29 skills and 4 workflow agents
appear flat. Same pattern gstack uses. Symlinking SKILL.md (not the
whole directory) keeps bundled references/ resolving correctly because
paths resolve relative to the symlink target's real location.

The installer is idempotent (safe to re-run after `git pull`) and
`--uninstall` only removes symlinks that point back into the repo.

Update README install, update, and uninstall sections accordingly.
Named `create` rather than `setup` to avoid confusion with `/setup-cs`.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Ignore local .claude/ settings directory

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Ignore .cursor/ and .env files

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* Rename /setup-cs to /setup-brain

The previous name was picked to avoid colliding with the generic
/setup command used by other skill stacks, but it was terse and
unmemorable. /setup-brain ties directly to the product's central
metaphor — every other skill reads from and writes to "the Brain"
(the Context Vault at ~/.creativestack/), so naming the onboarding
skill after the thing it creates is self-documenting.

Also avoids confusion with the repo's new ./create installer, where
the prior pairing of ./setup (script) and /setup-cs (skill) would
have been needlessly similar if the installer had been called setup.

- Rename _build/templates/setup-cs → setup-brain
- Rename skills/setup-cs → setup-brain (regenerated from template)
- Replace all /setup-cs references across templates, shared content,
  README, CLAUDE.md, CHANGELOG, scripts/check-brain.sh, and create
- Rebuild all skills to propagate the rename

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: camawjones

.gitignore

@@ -11,6 +11,12 @@ Thumbs.db
 # IDE
 .vscode/
 .idea/
+.cursor/
 
 # Local Claude Code settings
 .claude/
+
+# Environment variables
+.env
+.env.local
+.env.*.local

CHANGELOG.md

@@ -7,7 +7,7 @@ The format is based on Keep a Changelog, and this project adheres to Semantic Ve
 ## [Unreleased]
 
 ### Changed
-- Converted from Claude Code plugin to flat skill stack. Skills now invoked as `/skill-name` instead of `/creativestack:skill-name`. The `setup` skill was renamed to `setup-cs` to avoid collision with other stacks.
+- Converted from Claude Code plugin to flat skill stack. Skills now invoked as `/skill-name` instead of `/creativestack:skill-name`. The `setup` skill was renamed to `setup-brain` to avoid collision with other stacks.
 - Removed `.claude-plugin/` manifest and `hooks/` directory (plugin-only).
 - `scripts/check-brain.sh` self-locates instead of relying on `$CLAUDE_PLUGIN_ROOT`; install manually as a SessionStart hook via `~/.claude/settings.json` if desired.
 

CLAUDE.md

@@ -4,7 +4,7 @@ AI skill suite for creative professionals. 29 skills across 10 categories, plus
 
 ## Project structure
 
-This is a flat Claude Code skill stack installed at `~/.claude/skills/creativestack`. Skills are invoked flat as `/skill-name` (the parent directory is just a grouping — it does not create a namespace). The `setup` skill is named `setup-cs` to avoid collision with other stacks.
+This is a flat Claude Code skill stack installed at `~/.claude/skills/creativestack`. Skills are invoked flat as `/skill-name` (the parent directory is just a grouping — it does not create a namespace). The `setup` skill is named `setup-brain` to avoid collision with other stacks.
 
 - Skills live in `skills/*/SKILL.md` (generated — do not edit directly)
 - Source templates live in `_build/templates/*/SKILL.md.tmpl`

README.md

@@ -42,7 +42,7 @@ cd ~/.claude/skills/creativestack
 
 `./create` symlinks each skill's `SKILL.md` into `~/.claude/skills/<name>/SKILL.md` and each workflow agent into `~/.claude/agents/<name>.md`, so Claude Code discovers them as flat top-level entries. Restart Claude Code and all 29 skills and 4 agents are available — invoked flat as `/creative-brief`, `/meeting-notes`, `/competitor-audit`, and so on.
 
-The one skill with a non-obvious name is `/setup-cs` (renamed from `/setup` to avoid colliding with other skill stacks). Everything else matches its directory name.
+The one skill with a non-obvious name is `/setup-brain` (renamed from `/setup` to avoid colliding with other skill stacks). Everything else matches its directory name.
 
 ### Updates
 
@@ -83,10 +83,10 @@ Skip this and everything still works — each skill does its own brain-freshness
 **1. Set up your brain (5 minutes - optional but transformative)**
 
 ```
-/setup-cs
+/setup-brain
 ```
 
-This is an onboarding flow, not an installer — the skills are already installed by the `git clone`. `/setup-cs` walks you through who you are, your team, your tone of voice, and your standard methodology, and creates your Context Vault at `~/.creativestack/`. Every other skill reads from this vault and gets sharper because of it.
+This is an onboarding flow, not an installer — the skills are already installed by the `git clone`. `/setup-brain` walks you through who you are, your team, your tone of voice, and your standard methodology, and creates your Context Vault at `~/.creativestack/`. Every other skill reads from this vault and gets sharper because of it.
 
 You can skip this entirely and use any skill immediately — they'll work, just generically.
 
@@ -123,7 +123,7 @@ Every CreativeStack skill reads from and (with permission) writes to `~/.creativ
 
 ### Foundation files
 
-Created by `/setup-cs` - the baseline:
+Created by `/setup-brain` - the baseline:
 
 | File | What it holds |
 |---|---|
@@ -225,7 +225,7 @@ These all live in the brain precisely so you don't have to touch the skill templ
 
 | Skill | What it does |
 |---|---|
-| `/setup-cs` | Initialise or refresh your brain - profile, tone of voice, team, clients, methodology |
+| `/setup-brain` | Initialise or refresh your brain - profile, tone of voice, team, clients, methodology |
 | `/update-voice` | Refine your tone-of-voice guidelines with examples and do/don't rules |
 
 ### Daily operations

_build/build-skills.sh

@@ -33,7 +33,7 @@ for tmpl in "$SCRIPT_DIR"/templates/*/SKILL.md.tmpl; do
   # Replace {{PROJECT_STATE}} placeholder (only present in project-aware skills)
   content="${content//\{\{PROJECT_STATE\}\}/$PROJECT_STATE}"
 
-  # Replace {{VISUAL_STYLE_DEFAULT}} placeholder (used by setup-cs skill to seed brain)
+  # Replace {{VISUAL_STYLE_DEFAULT}} placeholder (used by setup-brain skill to seed brain)
   content="${content//\{\{VISUAL_STYLE_DEFAULT\}\}/$VISUAL_STYLE_DEFAULT}"
 
   # Write generated SKILL.md to skills/ directory

_build/shared/preamble.md

@@ -11,7 +11,7 @@ if [ -n "$BRAIN_DIR" ]; then
   echo "BRAIN: $BRAIN_DIR"
   ls "$BRAIN_DIR"/*.md 2>/dev/null | while read f; do echo "  $(basename "$f")"; done
 else
-  echo "BRAIN: not configured (run /setup-cs to set up)"
+  echo "BRAIN: not configured (run /setup-brain to set up)"
 fi
 ```
 
@@ -20,7 +20,7 @@ Use the content to inform and contextualize all outputs. If the brain doesn't ex
 proceed generically — the skill still works, just without your specific context.
 
 When the brain is not configured, mention once at the end of output:
-"Tip: Run /setup-cs to add your context — skills produce better results with it."
+"Tip: Run /setup-brain to add your context — skills produce better results with it."
 
 ### Brain Freshness Check
 
@@ -41,14 +41,14 @@ to refresh — lightly, not annoyingly.
 |---|---|---|
 | `learnings.md` | 60 days | continuous skill use — skills append as they run |
 | `case-studies.md` | 90 days | `/case-study` |
-| `clients.md` | 90 days | `/setup-cs` |
+| `clients.md` | 90 days | `/setup-brain` |
 | `team.md` | 90 days | `/resource-conflict` Setup team mode |
 | `freelance-bench.md` | 120 days | `/resource-conflict` Setup bench mode |
 | `rate-card.md` | 180 days | `/project-profitability` Setup rates mode |
-| `methodology.md` | 180 days | `/setup-cs` |
+| `methodology.md` | 180 days | `/setup-brain` |
 | `tone-of-voice.md` | 180 days | `/update-voice` |
 | `sow-style.md` | 180 days | `/sow-generator` Edit style mode |
-| `profile.md` | 365 days | `/setup-cs` |
+| `profile.md` | 365 days | `/setup-brain` |
 | `projects/*` | N/A | living documents — updated by their own skills |
 | `prospects/*` | 90 days | `/pitch-research` Refresh mode |
 
@@ -70,7 +70,7 @@ Keep it to 2-3 lines maximum. If more than 3 files are stale, summarise:
 ```
 ---
 📅 **Brain freshness:** {N} brain files are stale ({list names briefly}). Consider a
-session of `/setup-cs` Refresh mode to bring everything current.
+session of `/setup-brain` Refresh mode to bring everything current.
 ```
 
 5. **Severity gating:** only surface the check if at least one file is
@@ -85,7 +85,7 @@ session of `/setup-cs` Refresh mode to bring everything current.
    because brain data is stale. Surface, then proceed.
 
 8. **No brain, no check:** if the brain isn't configured at all, skip the
-   freshness check entirely. The `/setup-cs` nudge from the Brain
+   freshness check entirely. The `/setup-brain` nudge from the Brain
    Discovery step is enough.
 
 This check is lightweight by design. The goal is a gentle reminder, not an

_build/shared/visual-style-default.md

@@ -180,4 +180,4 @@ still looks good in three months. A rainbow deliverable looks dated by Friday.
 ---
 
 *This file is read by CreativeStack skills that produce visual output. Edit it to match
-your house style. Run `/setup-cs` to regenerate it from a preset.*
+your house style. Run `/setup-brain` to regenerate it from a preset.*

_build/templates/ai-audit/SKILL.md.tmpl

@@ -1,6 +1,6 @@
 ---
 name: ai-audit
-description: AI readiness audit for creative professionals — four modes. Snapshot (the default — 10-minute self-assessment producing maturity scores across 5 sharpened categories, top 3 opportunities, 3-5 curated quick wins mapped to real CreativeStack skills, protected zone scoring, ROI framing in saved hours/revenue, Slack summary, discipline-aware questions, AI-advanced user branching). Leadership Pack (restructures the same data for a CEO/board audience with benchmark comparison, executive one-pager, and risk/opportunity framing so the user can take it to leadership to justify budget or hiring). Team Workshop (longer intake from multiple team member perspectives, surfaces disagreements as a distinct output, produces a workshop-ready slide outline plus the bringing-your-team-along change management plan with team anxieties, verbatim framings, protected-zone statement, 30-60-90 day plan, and handling-the-skeptic guidance). Deep Audit (includes prior-audit delta tracking, workflow walkthrough, tool stack review against tools.md). Uses practitioner-view industry benchmarks segmented by discipline and team size so users can answer "3/5 — is that good?". Writes audits to the brain as dated artefacts so repeat runs show measurable progress. Opt-in writes to learnings.md § AI maturity and tools.md. Every opportunity is mapped to a specific CreativeStack skill from a curated library, never invented. Use when a creative reaches for "where does AI fit in my practice", "I need to present AI strategy to my CEO", "I'm losing pitches to AI-native competitors", "which CreativeStack skills matter most for me", or "our team is nervous about AI and I need to lead on this". Also use as the natural second skill after /setup-cs.
+description: AI readiness audit for creative professionals — four modes. Snapshot (the default — 10-minute self-assessment producing maturity scores across 5 sharpened categories, top 3 opportunities, 3-5 curated quick wins mapped to real CreativeStack skills, protected zone scoring, ROI framing in saved hours/revenue, Slack summary, discipline-aware questions, AI-advanced user branching). Leadership Pack (restructures the same data for a CEO/board audience with benchmark comparison, executive one-pager, and risk/opportunity framing so the user can take it to leadership to justify budget or hiring). Team Workshop (longer intake from multiple team member perspectives, surfaces disagreements as a distinct output, produces a workshop-ready slide outline plus the bringing-your-team-along change management plan with team anxieties, verbatim framings, protected-zone statement, 30-60-90 day plan, and handling-the-skeptic guidance). Deep Audit (includes prior-audit delta tracking, workflow walkthrough, tool stack review against tools.md). Uses practitioner-view industry benchmarks segmented by discipline and team size so users can answer "3/5 — is that good?". Writes audits to the brain as dated artefacts so repeat runs show measurable progress. Opt-in writes to learnings.md § AI maturity and tools.md. Every opportunity is mapped to a specific CreativeStack skill from a curated library, never invented. Use when a creative reaches for "where does AI fit in my practice", "I need to present AI strategy to my CEO", "I'm losing pitches to AI-native competitors", "which CreativeStack skills matter most for me", or "our team is nervous about AI and I need to lead on this". Also use as the natural second skill after /setup-brain.
 ---
 
 {{PREAMBLE}}
@@ -139,7 +139,7 @@ point. But if any of these ran earlier, pull in:
 
 | Skill run earlier | What to use |
 |---|---|
-| `/setup-cs` | Skip the type/discipline/size questions |
+| `/setup-brain` | Skip the type/discipline/size questions |
 | `/project-profitability` | Skip commercials intake; use the data |
 | `/timesheet-summary` | Skip bottleneck question; use the findings |
 | `/resource-conflict` | Skip team capacity questions |
@@ -166,7 +166,7 @@ Report brain depth:
 Or thin:
 
 > "Brain depth: Minimal — no brain configured. I'll run the audit
-> with neutral language and recommend `/setup-cs` at
+> with neutral language and recommend `/setup-brain` at
 > the end so the next audit compounds."
 
 ### Step 4: Intake
@@ -376,7 +376,7 @@ Offer **at most one** brain enrichment or next step. Pick the most
 valuable:
 
 - **No brain configured** → "The biggest lift you can give this
-  skill is configuring `/setup-cs` first. Takes 5
+  skill is configuring `/setup-brain` first. Takes 5
   minutes and every future audit will be dramatically sharper."
 - **Prior audit exists and delta is positive** → "You moved from
   {X} to {Y} overall — measurable progress. This is exactly the
@@ -701,7 +701,7 @@ time.
 
 - **No brain at all** → use neutral language ("your work", "your
   clients"), skip type-specific adaptations, end with a strong
-  nudge to `/setup-cs` for next time.
+  nudge to `/setup-brain` for next time.
 - **Brain exists but very thin** (profile only) → run the audit
   but note reduced depth in the output header.
 - **User is AI-advanced** → branch per Step 5. Never patronise.
@@ -733,15 +733,15 @@ time.
 
 ### Upstream (what feeds ai-audit)
 
-- **No brain** → "`/setup-cs` first — takes 5 minutes and every future audit will be sharper."
+- **No brain** → "`/setup-brain` first — takes 5 minutes and every future audit will be sharper."
 - **No tone of voice** → "Consider `/update-voice` before the next audit so the output matches your voice."
 - **Running as a follow-up to setup** → natural flow, no nudge needed.
 
 ### Downstream (what ai-audit triggers)
 
 Based on the audit's findings, suggest 1-2 of:
 
-- **Lowest category = Knowledge & IP** → "Start with `/case-study` and `/setup-cs` — institutional memory is the single highest-leverage move."
+- **Lowest category = Knowledge & IP** → "Start with `/case-study` and `/setup-brain` — institutional memory is the single highest-leverage move."
 - **Lowest category = Ops & Admin** → "Run `/sow-generator` or `/timesheet-summary` this week — fastest wins, lowest risk."
 - **Lowest category = Creative Workflow** → "Run `/brief-sharpener` on your next incoming brief to feel the difference immediately."
 - **Lowest category = Client Comms** → "Run `/status-update` this week — 5-minute drafts your clients will notice."

_build/templates/ai-audit/references/quick-wins.md

@@ -260,7 +260,7 @@ Each win has this schema:
 - **What it looks like:** Configure your CreativeStack Brain — profile, team, tone of voice, clients, and tools — so every skill produces personalised output
 - **Time to value:** First value this week
 - **Effort:** L
-- **Skill:** `/setup-cs`
+- **Skill:** `/setup-brain`
 - **User types:** all
 - **Disciplines:** all
 - **Tier:** beginner-only

_build/templates/ai-audit/references/rubric.md

@@ -115,7 +115,7 @@ fed back into briefs, pitches, and proposals automatically.
   curate)
 - Personal craft rituals
 
-**Maps to skills:** `/case-study`, `/post-mortem`, `/update-voice`, `/setup-cs`, `/brand-guidelines`
+**Maps to skills:** `/case-study`, `/post-mortem`, `/update-voice`, `/setup-brain`, `/brand-guidelines`
 
 ---
 
@@ -325,7 +325,7 @@ answers or the brain has no context), the scoring is directional
 only. Flag with:
 
 > **Confidence:** Low — based on brief answers without brain
-> context. Treat scores as directional. Running `/setup-cs`
+> context. Treat scores as directional. Running `/setup-brain`
 > first would sharpen the next audit significantly.
 
 Never score confidently off thin evidence.