qualia-framework 2.6.0 → 3.2.0

package/framework/skills/qualia-framework-audit/SKILL.md

@@ -1,604 +0,0 @@
----
-name: qualia-framework-audit
-description: "Deep audit of the Qualia framework itself — CLAUDE.md, agents, skills, hooks, knowledge, rules, and overall architecture. Spawns 6 specialist agents to review every dimension and produces a unified report. Use when you want to audit the framework, find redundancies, gaps, or inconsistencies. Trigger on 'framework audit', 'audit framework', 'review framework', 'qualia-framework-audit'."
-tags: [framework, audit, meta, agents, skills, hooks]
----
-
-# Qualia Framework Audit
-
-Deep parallel-agent audit of the Qualia framework infrastructure (`~/.claude/`). Not a project review — a **meta-review** of the tools, agents, skills, hooks, and knowledge that power every project.
-
-## Usage
-
-- `/qualia-framework-audit` — Full audit (all 6 dimensions)
-- `/qualia-framework-audit --agents` — Agents & teams only
-- `/qualia-framework-audit --skills` — Skill portfolio only
-- `/qualia-framework-audit --hooks` — Hook infrastructure only
-- `/qualia-framework-audit --knowledge` — Knowledge base only
-- `/qualia-framework-audit --config` — CLAUDE.md, rules, settings only
-- `/qualia-framework-audit --coherence` — Cross-cutting architecture only
-
-## Process
-
-### Step 1: Parse Arguments
-
-Extract mode from $ARGUMENTS. Default to `full`.
-
-Modes: `full`, `agents`, `skills`, `hooks`, `knowledge`, `config`, `coherence`.
-
-### Step 2: Inventory the Framework
-
-Run these to build the current inventory:
-
-```bash
-# Count everything
-echo "=== INVENTORY ==="
-echo "Agents: $(ls ~/.claude/agents/*.md 2>/dev/null | wc -l)"
-echo "Teams: $(ls ~/.claude/agents/teams/*.md 2>/dev/null | wc -l)"
-echo "Skills: $(ls -d ~/.claude/skills/*/ 2>/dev/null | wc -l)"
-echo "Hooks: $(ls ~/.claude/hooks/*.sh 2>/dev/null | wc -l)"
-echo "Knowledge: $(ls ~/.claude/knowledge/*.md 2>/dev/null | wc -l)"
-echo "Rules: $(ls ~/.claude/rules/*.md 2>/dev/null | wc -l)"
-echo "Archive agents: $(ls ~/.claude/archive/agents/*.md 2>/dev/null | wc -l)"
-echo "Archive skills: $(find ~/.claude/archive/skills* -name 'SKILL.md' 2>/dev/null | wc -l)"
-```
-
-```bash
-# Get all agent names
-ls ~/.claude/agents/*.md | xargs -I{} basename {} .md | sort
-```
-
-```bash
-# Get all skill names
-ls -d ~/.claude/skills/*/ | xargs -I{} basename {} | sort
-```
-
-```bash
-# Get all hook names
-ls ~/.claude/hooks/*.sh | xargs -I{} basename {} .sh | sort
-```
-
-```bash
-# Get settings.json hook registrations
-node -e "const s=require('$HOME/.claude/settings.json');Object.entries(s.hooks||{}).forEach(([event,matchers])=>{matchers.forEach(m=>{m.hooks.forEach(h=>{console.log(event+' ['+m.matcher+'] → '+h.command.split('/').pop())})})})"
-```
-
-Read CLAUDE.md, all rules, settings.json. Store everything — inline into agent prompts.
-
-### Step 3: Spawn Agents
-
-Based on mode, spawn agents in a **single message** with multiple Agent() calls.
-
-| Mode | Agents Spawned |
-|------|---------------|
-| `full` | All 6 (parallel) |
-| `agents` | Agent 2 only |
-| `skills` | Agent 3 only |
-| `hooks` | Agent 4 only |
-| `knowledge` | Agent 5 only |
-| `config` | Agent 1 only |
-| `coherence` | Agent 6 only |
-
-**CRITICAL**: Inline all inventory data and file contents into each agent prompt. `@` references don't work across Agent() boundaries.
-
----
-
-#### Agent 1: Core Configuration Reviewer
-
-```
-Agent(
-  prompt="You are auditing the core configuration of a Claude Code framework.
-
-<claude_md>
-{CLAUDE.md content}
-</claude_md>
-
-<rules>
-{All rules/*.md content}
-</rules>
-
-<settings>
-{settings.json content}
-</settings>
-
-<task>
-Review the core configuration for quality, completeness, and consistency.
-
-1. **CLAUDE.md Quality**
-   - Is the identity section clear and complete?
-   - Are all rules actionable (not vague)?
-   - Does the workflow section match actual behavior?
-   - Are there contradictions between sections?
-   - Is the compaction section preserving the right things?
-   - Is the Qualia Mode section accurate to what the framework actually does?
-
-2. **Rules Consistency**
-   - Do rules/*.md files contradict each other?
-   - Do rules contradict CLAUDE.md?
-   - Are any rules redundant (same thing said in multiple places)?
-   - Are there gaps — common scenarios with no rule?
-   - Are rules specific enough to be enforceable?
-
-3. **Settings.json Correctness**
-   - Are all hook files referenced in settings.json actually present on disk?
-   - Are hook matchers correct (right tool names)?
-   - Are permissions sensible (allow/deny)?
-   - Are MCP servers all needed and correctly configured?
-   - Are environment variables appropriate?
-
-4. **Cross-References**
-   - Does CLAUDE.md reference files that exist?
-   - Do 'See rules/X.md' references point to real files?
-   - Are knowledge file references in MEMORY.md still valid?
-
-For EVERY finding:
-- **What**: description
-- **Where**: file:line or file
-- **Why**: impact (confusion, inconsistency, broken behavior)
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Core config review"
-)
-```
-
-#### Agent 2: Agent & Team Architecture Reviewer
-
-```
-Agent(
-  prompt="You are auditing all agent definitions and team templates in a Claude Code framework.
-
-<inventory>
-Agents: {list of all agent .md files}
-Teams: {list of all team .md files}
-</inventory>
-
-Your job: Read EVERY agent file in ~/.claude/agents/ and every team file in ~/.claude/agents/teams/. Then analyze.
-
-<task>
-1. **Agent Quality**
-   - Read each agent .md file
-   - Is the description accurate to what the agent actually does?
-   - Is the system prompt clear and actionable?
-   - Are tool permissions appropriate (not too broad, not too narrow)?
-   - Is the model setting correct (inherit vs specific)?
-
-2. **Agent Overlap**
-   - Which agents have overlapping responsibilities?
-   - Could any agents be merged without losing capability?
-   - Example overlaps to check: code-simplicity-reviewer vs architecture-strategist, qualia-verifier vs qualia-plan-checker
-
-3. **Agent Gaps**
-   - What common tasks have no dedicated agent?
-   - Are there agents that skills reference but don't exist?
-   - Are there agents that nothing references (orphans)?
-
-4. **Team Template Quality**
-   - Do team templates reference agents that exist?
-   - Are orchestration patterns correct (fan-out, pipeline, review-loop)?
-   - Are shared context lists pointing to real files?
-   - Are coordination rules clear?
-
-5. **Naming & Consistency**
-   - Is naming consistent (qualia-* prefix for pipeline agents, descriptive names for specialists)?
-   - Are frontmatter fields consistent across agents?
-   - Do descriptions follow the same structure?
-
-For EVERY finding:
-- **What**: description
-- **Where**: file
-- **Why**: impact
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Agent architecture review"
-)
-```
-
-#### Agent 3: Skill Portfolio Reviewer
-
-```
-Agent(
-  prompt="You are auditing all skills in a Claude Code framework. There are 65+ skills.
-
-<inventory>
-Skills: {list of all skill directory names}
-</inventory>
-
-Your job: Read the SKILL.md from each skill directory in ~/.claude/skills/. Then analyze.
-
-<task>
-1. **Skill Quality** (sample 15-20 representative skills across categories)
-   - Is the description accurate and under 100 words (INFR-03)?
-   - Are trigger phrases comprehensive?
-   - Is the process well-structured and actionable?
-   - Does it reference agents/tools that exist?
-
-2. **Skill Overlap**
-   - Which skills do similar things? Map overlaps.
-   - Key overlaps to check:
-     - qualia-review vs deep-research vs qualia-production-check vs qualia-optimize
-     - deploy vs ship vs deploy-verify
-     - qualia-debug vs qualia-optimize --fix
-     - critique vs polish vs bolder vs design-quieter vs distill
-   - For each overlap: is the distinction clear? Would a user know which to use?
-
-3. **Skill Gaps**
-   - What common workflows have no skill?
-   - Are there skills in the archive that should be active?
-   - Check archive: ~/.claude/archive/skills-2026-03/ and ~/.claude/archive/skills/
-
-4. **Skill-Agent Alignment**
-   - Do skills that spawn agents reference correct subagent_type values?
-   - Do skills reference agent capabilities that match agent definitions?
-
-5. **Category Health**
-   - Group skills by category (pipeline, design, deploy, research, domain)
-   - Which categories are overcrowded?
-   - Which are underdeveloped?
-
-6. **Dead Skill Detection**
-   - Are any skills never referenced in CLAUDE.md, other skills, or hooks?
-   - Are any skill triggers so narrow they'd never fire?
-
-For EVERY finding:
-- **What**: description
-- **Where**: skill name or file
-- **Why**: impact (user confusion, wasted context, broken workflow)
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Skill portfolio review"
-)
-```
-
-#### Agent 4: Hook Infrastructure Reviewer
-
-```
-Agent(
-  prompt="You are auditing all hooks in a Claude Code framework.
-
-<settings_hooks>
-{Hook registrations from settings.json}
-</settings_hooks>
-
-<hook_files>
-{List of all .sh files in hooks/}
-</hook_files>
-
-Your job: Read EVERY hook .sh file in ~/.claude/hooks/ and cross-reference with settings.json. Then analyze.
-
-<task>
-1. **Hook Correctness**
-   - Read each hook file
-   - Does it handle stdin JSON parsing correctly?
-   - Does it produce valid JSON stdout for Claude?
-   - Does error handling cover edge cases (missing files, missing commands)?
-   - Does it use correct exit codes (0 = allow, 2 = block)?
-
-2. **Hook-Settings Alignment**
-   - Is every hook .sh file registered in settings.json?
-   - Is every settings.json hook reference pointing to an existing file?
-   - Are matchers correct (e.g., 'Bash' for git commands, 'Edit'/'Write' for file changes)?
-   - Are event types correct (PreToolUse, PostToolUse, etc.)?
-
-3. **Hook Performance**
-   - Are there hooks that could be slow (network calls, builds, long file scans)?
-   - Do hooks have early-exit guards for irrelevant commands?
-   - Is the total hook chain likely to stay under 500ms (INFR-02)?
-
-4. **Branding Consistency**
-   - Do all hooks source qualia-colors.sh?
-   - Do all hooks use q_header, q_pass, q_fail, q_warn, q_skip functions?
-   - Do blocking hooks include actionable systemMessage telling Claude to FIX?
-
-5. **Coverage Gaps**
-   - Are there dangerous operations without hook protection?
-   - Are there tool events without any hooks?
-   - Should any PostToolUse hooks be PreToolUse (or vice versa)?
-
-6. **Orphan Detection**
-   - Any hook files not registered in settings.json?
-   - Any hooks that do nothing (empty/stub implementations)?
-
-For EVERY finding:
-- **What**: description
-- **Where**: hook file or settings.json section
-- **Why**: impact (missed protection, performance, broken enforcement)
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Hook infrastructure review"
-)
-```
-
-#### Agent 5: Knowledge Base & Memory Reviewer
-
-```
-Agent(
-  prompt="You are auditing the knowledge base and memory system of a Claude Code framework.
-
-<knowledge_files>
-{List of all files in knowledge/}
-</knowledge_files>
-
-<memory_index>
-{MEMORY.md content}
-</memory_index>
-
-Your job: Read EVERY .md file in ~/.claude/knowledge/ and the memory index. Then analyze.
-
-<task>
-1. **Knowledge Staleness**
-   - Check modification dates of each knowledge file
-   - Flag files older than 30 days
-   - Check if content references outdated project states, URLs, or configurations
-   - Cross-reference with qualia-context.md (the source of truth for projects)
-
-2. **Knowledge Accuracy**
-   - Read qualia-context.md: are project statuses current? Any missing projects?
-   - Read deployment-map.md: do deploy commands match reality?
-   - Read supabase-patterns.md: are patterns still current best practices?
-   - Read voice-agent-patterns.md: are provider details current?
-   - Read common-fixes.md: are fixes still relevant?
-   - Read learned-patterns.md: are learned rules still applicable?
-   - Read client-prefs.md: any missing client preferences?
-   - Read employees.md: is team info current?
-
-3. **Knowledge Completeness**
-   - Are there knowledge gaps? Topics that come up frequently but have no file?
-   - Potential gaps: pricing/billing patterns, testing patterns, error monitoring patterns
-   - Are subdirectories (project-notes/, retros/) populated and useful?
-
-4. **Memory System Health**
-   - Is MEMORY.md well-organized?
-   - Are entries concise (under 150 chars)?
-   - Are there stale or outdated entries?
-   - Do file references in MEMORY.md point to existing files?
-   - Is the Knowledge Routing section accurate?
-
-5. **Cross-Reference Integrity**
-   - Do CLAUDE.md references to knowledge files point to existing files?
-   - Do skills that reference knowledge files point to existing files?
-   - Are there circular or broken references?
-
-6. **Duplicate Content**
-   - Is the same information stored in multiple knowledge files?
-   - Could any knowledge files be merged?
-
-For EVERY finding:
-- **What**: description
-- **Where**: file
-- **Why**: impact (stale info leads to wrong decisions, missing info leads to repeated mistakes)
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Knowledge base review"
-)
-```
-
-#### Agent 6: Framework Coherence & Architecture Reviewer
-
-```
-Agent(
-  prompt="You are conducting a cross-cutting architectural review of a Claude Code framework.
-
-<inventory>
-{Full directory tree of ~/.claude (top 2 levels)}
-Agents: {count}
-Skills: {count}
-Hooks: {count}
-Knowledge: {count}
-Archive agents: {count}
-Archive skills: {count}
-</inventory>
-
-<claude_md>
-{CLAUDE.md content}
-</claude_md>
-
-Your job: Review the overall architecture, organization, and coherence of the framework. Read key files across all dimensions.
-
-<task>
-1. **File Organization**
-   - Is the directory structure logical and consistent?
-   - Are files in the right directories?
-   - Are naming conventions consistent (kebab-case, qualia-* prefix)?
-   - Is the archive/ directory properly maintained?
-   - Are there files at the root level that should be in subdirectories?
-
-2. **Cross-Component Alignment**
-   - Do skills reference agents that exist with correct subagent_type values?
-   - Do team templates reference agents that exist?
-   - Do hooks reference paths that exist?
-   - Does settings.json match the actual file structure?
-   - Does MEMORY.md route to files that exist?
-
-3. **Dead Code & Cruft**
-   - Are there files in the archive that should be deleted entirely?
-   - Are there active files that are effectively dead (never referenced)?
-   - Are there config files for removed features?
-   - Check: cache/, backups/, paste-cache/, shell-snapshots/ — are retention policies working?
-
-4. **Naming Consistency**
-   - Agent names: consistent pattern?
-   - Skill names: consistent pattern?
-   - Hook names: consistent pattern?
-   - Knowledge file names: consistent pattern?
-
-5. **Documentation Quality**
-   - Does the framework have adequate documentation for onboarding (if someone else needs to understand it)?
-   - Are the install scripts (install.sh, install.ps1) current?
-   - Is MCP_SETUP.md current?
-   - Is cowork-knowledge-pack.md useful or cruft?
-
-6. **Scalability Concerns**
-   - With 65+ skills, is the skill discovery/loading efficient?
-   - Are there performance concerns with the current architecture?
-   - What happens when the number of projects/agents/skills doubles?
-
-7. **Security**
-   - Is .credentials.json properly protected (in deny list)?
-   - Is .env.claude properly protected?
-   - Are there any secrets exposed in non-protected files?
-   - Is .gitignore comprehensive?
-   - Is .claudeignore comprehensive?
-
-For EVERY finding:
-- **What**: description
-- **Where**: file or directory
-- **Why**: impact (maintenance burden, confusion, security risk)
-- **Fix**: concrete suggestion
-- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
-</task>",
-  subagent_type="general-purpose",
-  description="Framework coherence review"
-)
-```
-
-### Step 4: Collect & Score
-
-After all agents return:
-
-1. **Deduplicate** — Same file from multiple agents: keep the most detailed finding
-2. **Categorize** by dimension:
-   - Core Config (Agent 1)
-   - Agents & Teams (Agent 2)
-   - Skills (Agent 3)
-   - Hooks (Agent 4)
-   - Knowledge (Agent 5)
-   - Architecture (Agent 6)
-3. **Sort by severity**: CRITICAL > HIGH > MEDIUM > LOW
-4. **Count totals** per severity and per dimension
-
-### Step 5: Present Report
-
-Output as direct text (NOT via Bash):
-
-```
-◆ FRAMEWORK AUDIT — QUALIA INFRASTRUCTURE REVIEW
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-  Date         {date}
-  Scope        {full / specific dimension}
-  Framework    v{qualia-framework VERSION}
-
-  INVENTORY
-  Agents       {N}     Teams     {N}
-  Skills       {N}     Hooks     {N}
-  Knowledge    {N}     Rules     {N}
-
-  FINDINGS
-  CRITICAL     {N}     HIGH      {N}
-  MEDIUM       {N}     LOW       {N}
-
-── CRITICAL (must fix) ─────────────────────────
-  {findings or "None"}
-
-── HIGH (should fix) ───────────────────────────
-  {findings}
-
-── MEDIUM (recommended) ────────────────────────
-  {findings}
-
-── LOW (nice to have) ──────────────────────────
-  {findings}
-
-── DIMENSION HEALTH ────────────────────────────
-
-  Core Config     {PASS/ISSUES}  ({N} findings)
-  Agents & Teams  {PASS/ISSUES}  ({N} findings)
-  Skills          {PASS/ISSUES}  ({N} findings)
-  Hooks           {PASS/ISSUES}  ({N} findings)
-  Knowledge       {PASS/ISSUES}  ({N} findings)
-  Architecture    {PASS/ISSUES}  ({N} findings)
-
-── OVERLAP MAP ─────────────────────────────────
-  {skill/agent overlaps identified}
-
-── GAP ANALYSIS ────────────────────────────────
-  {missing capabilities identified}
-
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-```
-
-### Step 6: Save Report
-
-Write to `~/.claude/.planning/FRAMEWORK-AUDIT.md`:
-
-```markdown
----
-date: YYYY-MM-DD HH:MM
-scope: full|agents|skills|hooks|knowledge|config|coherence
-critical: N
-high: N
-medium: N
-low: N
-agents_count: N
-skills_count: N
-hooks_count: N
-status: healthy|needs_attention|critical_issues
----
-
-# Framework Audit — YYYY-MM-DD
-
-## Executive Summary
-{2-3 sentence overview}
-
-## Findings by Dimension
-
-### Core Configuration
-{Agent 1 findings}
-
-### Agents & Teams
-{Agent 2 findings}
-
-### Skill Portfolio
-{Agent 3 findings}
-
-### Hook Infrastructure
-{Agent 4 findings}
-
-### Knowledge Base
-{Agent 5 findings}
-
-### Architecture & Coherence
-{Agent 6 findings}
-
-## Overlap Map
-{Table of overlapping skills/agents with recommendation}
-
-## Gap Analysis
-{Missing capabilities with priority}
-
-## Action Items
-| # | Severity | Dimension | Finding | Fix |
-|---|----------|-----------|---------|-----|
-| 1 | CRITICAL | ... | ... | ... |
-| 2 | HIGH | ... | ... | ... |
-```
-
-### Step 7: Offer Next Steps
-
-```
-Framework audit complete. Options:
-
-1. Fix critical/high findings now
-2. Create a fix milestone: /qualia-new-milestone
-3. Deep dive on a specific dimension: /qualia-framework-audit --{dimension}
-4. Review full report: cat ~/.claude/.planning/FRAMEWORK-AUDIT.md
-```
-
-## Key Principles
-
-- **Parallel Execution**: Always spawn all 6 agents in ONE message for speed
-- **Evidence-Based**: Every finding must reference specific files
-- **Actionable**: Every recommendation must be implementable
-- **Prioritized**: CRITICAL > HIGH > MEDIUM > LOW
-- **Non-Destructive**: This skill reads and reports — it never modifies files
-- **Self-Aware**: This skill audits the framework that contains it

package/framework/skills/qualia-guide/SKILL.md

@@ -1,32 +0,0 @@
----
-name: qualia-guide
-description: "Show the developer guide — how the Qualia framework works, the full project flow, and the 10 commands that matter. Use this skill whenever the user says 'guide', 'how does this work', 'explain the framework', 'qualia guide', 'show me the flow', or is confused about how to use the framework."
-user-invocable: true
-args: []
----
-
-# Qualia Developer Guide
-
-Display the employee guide that explains how the full workflow works.
-
-## Process
-
-1. Read `~/.claude/qualia-framework/references/employee-guide.md`
-2. Output its contents directly — this IS the guide, formatted for terminal display
-3. If in a project with `.planning/`, also show current position:
-   - Read STATE.md
-   - Show which step of the flow they're currently on
-   - Highlight the next command to run
-
-## Output Enhancement
-
-After showing the guide, if in a project, add:
-
-```
-── YOUR CURRENT POSITION ───────────────
-  You are at: [phase/step name]
-  Next command: /[command]
-```
-
----
-> Stuck? Type `/qualia-idk` · Lost? Type `/qualia-help` · Done for today? Type `/qualia-report`