claude-all-hands 1.0.0 → 1.0.2

package/.claude/skills/skill-development/scripts/validate-skill.sh

@@ -1,144 +0,0 @@
-#!/bin/bash
-# Skill File Validator
-# Validates SKILL.md files for correct structure and content
-
-set -euo pipefail
-
-# Usage
-if [ $# -eq 0 ]; then
-  echo "Usage: $0 <path/to/SKILL.md>"
-  echo ""
-  echo "Validates skill file for:"
-  echo "  - YAML frontmatter structure"
-  echo "  - Required fields (name, description)"
-  echo "  - Field formats and constraints"
-  echo "  - Body word count"
-  echo "  - Progressive disclosure check"
-  exit 1
-fi
-
-SKILL_FILE="$1"
-error_count=0
-warning_count=0
-
-echo "Validating skill: $SKILL_FILE"
-echo ""
-
-# Check: File exists
-if [ ! -f "$SKILL_FILE" ]; then
-  echo "ERROR: File not found: $SKILL_FILE"
-  exit 1
-fi
-
-# Check: Starts with ---
-FIRST_LINE=$(head -1 "$SKILL_FILE")
-if [ "$FIRST_LINE" != "---" ]; then
-  echo "ERROR: File must start with YAML frontmatter (---)"
-  exit 1
-fi
-
-# Check: Has closing ---
-if ! tail -n +2 "$SKILL_FILE" | grep -q '^---$'; then
-  echo "ERROR: Frontmatter not closed (missing second ---)"
-  exit 1
-fi
-
-# Extract frontmatter and body
-FRONTMATTER=$(sed -n '/^---$/,/^---$/{ /^---$/d; p; }' "$SKILL_FILE")
-BODY=$(awk '/^---$/{i++; next} i>=2' "$SKILL_FILE")
-
-# Validate: name field
-NAME=$(echo "$FRONTMATTER" | grep '^name:' | sed 's/name: *//' | sed 's/^"\(.*\)"$/\1/' | sed "s/^'\(.*\)'$/\1/")
-
-if [ -z "$NAME" ]; then
-  echo "ERROR: Missing required field: name"
-  ((error_count++))
-else
-  # Name length (max 64 chars)
-  name_length=${#NAME}
-  if [ $name_length -gt 64 ]; then
-    echo "ERROR: name too long ($name_length chars, max 64)"
-    ((error_count++))
-  fi
-
-  # Name format: lowercase-hyphenated
-  if ! echo "$NAME" | grep -qE '^[a-z0-9]+(-[a-z0-9]+)*$'; then
-    echo "ERROR: name must be lowercase-hyphenated (kebab-case): '$NAME'"
-    ((error_count++))
-  fi
-fi
-
-# Validate: description field
-DESCRIPTION=$(echo "$FRONTMATTER" | grep '^description:' | sed 's/description: *//')
-
-if [ -z "$DESCRIPTION" ]; then
-  echo "ERROR: Missing required field: description"
-  ((error_count++))
-else
-  desc_length=${#DESCRIPTION}
-
-  # Max 1024 chars
-  if [ $desc_length -gt 1024 ]; then
-    echo "ERROR: description too long ($desc_length chars, max 1024)"
-    ((error_count++))
-  fi
-
-  # Warn >300 chars
-  if [ $desc_length -gt 300 ]; then
-    echo "WARN: description is long ($desc_length chars, recommend <300)"
-    ((warning_count++))
-  fi
-fi
-
-# Validate: body word count
-if [ -n "$BODY" ]; then
-  word_count=$(echo "$BODY" | wc -w | tr -d ' ')
-
-  if [ "$word_count" -gt 3000 ]; then
-    echo "ERROR: body too long ($word_count words, max 3000)"
-    ((error_count++))
-  elif [ "$word_count" -gt 2000 ]; then
-    echo "WARN: body is long ($word_count words, target 1500-2000)"
-    ((warning_count++))
-  fi
-
-  # Progressive disclosure check for long bodies
-  if [ "$word_count" -gt 1500 ]; then
-    SKILL_DIR=$(dirname "$SKILL_FILE")
-
-    # Check for references/ or examples/ subdirs
-    has_progressive=0
-    if [ -d "$SKILL_DIR/references" ] || [ -d "$SKILL_DIR/examples" ] || [ -d "$SKILL_DIR/docs" ]; then
-      has_progressive=1
-    fi
-
-    # Also check for any .md files besides SKILL.md
-    other_md_count=$(find "$SKILL_DIR" -maxdepth 2 -name "*.md" ! -name "SKILL.md" 2>/dev/null | wc -l | tr -d ' ')
-    if [ "$other_md_count" -gt 0 ]; then
-      has_progressive=1
-    fi
-
-    if [ $has_progressive -eq 0 ]; then
-      echo "WARN: Long body ($word_count words) without progressive disclosure"
-      echo "      Consider adding references/ or examples/ subdirectories"
-      ((warning_count++))
-    fi
-  fi
-else
-  echo "WARN: Body is empty"
-  ((warning_count++))
-fi
-
-echo ""
-echo "---"
-
-if [ $error_count -eq 0 ] && [ $warning_count -eq 0 ]; then
-  echo "OK: All checks passed"
-  exit 0
-elif [ $error_count -eq 0 ]; then
-  echo "OK: Passed with $warning_count warning(s)"
-  exit 0
-else
-  echo "FAIL: $error_count error(s), $warning_count warning(s)"
-  exit 1
-fi

package/.claude/skills/specialist-builder/SKILL.md

@@ -1,348 +0,0 @@
----
-name: specialist-builder
-description: Use when user asks to "create an agent", "add a subagent", or needs guidance on agent frontmatter, triggering examples, system prompts, tools, or colors.
----
-
-# Specialist Builder
-
-Build specialist agents tailored to repository domains.
-
-For initial research this is a great researouce of specialist plugins (which ship agents with their commands and skills) (see their agent.md files):
-**https://github.com/wshobson/agents/tree/main/plugins**
-
-## When Triggered
-
-Main agent calls curator with this skill when:
-- No existing specialists match user's prompt domain
-- User confirms they want to architect a specialist
-- User says "create an agent", "add an agent", "write a subagent"
-
-## Process
-
-### Inference Logic
-
-When curator invocation already specifies details, skip those questions:
-- Specialist name specified → skip name confirmation
-- Source directories specified → skip directory question
-- Specific domain described → skip option proposal, go direct to requirements
-
-### 1. Research Phase (required)
-
-Before proposing options, research the domain:
-- Use research-tools skill to query best practices for the inferred domain
-- Gather documentation URLs for the specialist's area of focus
-- Synthesize findings for system prompt inclusion
-
-**AskUserQuestion**: "What documentation should this specialist learn from?"
-- Option A: [Inferred docs based on domain]
-- Option B: I'll provide specific URLs
-- Option C: Skip documentation research
-
-### 2. Extract Repo Patterns
-
-Use repomix-extraction skill to understand existing patterns:
-
-**AskUserQuestion**: "Which directories contain patterns this specialist should learn?"
-- Option A: [Inferred directory from prompt context]
-- Option B: Browse repository (I'll suggest directories)
-- Option C: Skip (no repo patterns needed)
-- Option D: New directory (specialist for new feature set)
-
-If directories specified:
-1. Use repomix-extraction skill to pack and analyze
-2. Extract patterns for system prompt embedding
-
-### 3. Confirm Name
-
-Suggest specialist name based on domain/scope:
-
-**AskUserQuestion**: "[suggested-name]?"
-- Option A: Yes, use this name
-- Option B: I'll provide a different name
-
-### 4. Propose Options (if not already specified)
-
-Use **AskUserQuestion** to present specialist options:
-
-```
-Based on your prompt, these specialist types could help:
-
-1. [Domain] Specialist - [what it would handle]
-2. [Domain] Specialist - [what it would handle]
-3. Custom - describe your own
-```
-
-### 5. Gather Requirements
-
-After user selects, use **AskUserQuestion** for each:
-
-**Scope**: "What specific areas should this specialist cover?"
-- Option A: [narrow scope]
-- Option B: [broader scope]
-- Option C: Custom
-
-**Agent Pattern**: "What type of agent is this?"
-- Analysis (code review, security audit, research)
-- Generation (code, tests, docs)
-- Validation (linting, checking, verification)
-- Orchestration (multi-step workflows)
-
-**Skills**: "What skills should this specialist have access to?"
-- repomix-extraction (always included for pattern discovery)
-- research-tools (web search, documentation)
-- claude-code-patterns (Claude Code best practices)
-- Custom skill (will need to be built)
-
-**Tools**: "What tools does this specialist need?"
-- Read-only (Read, Glob, Grep) - for analysis/research
-- With Bash - for running commands
-- Full access - omit tools field to inherit all
-
-### 6. Generate Agent Definition
-
-Return to main agent with proposed agent file following the structure below.
-
-**Pattern Embedding**: If repo patterns extracted, include in system prompt:
-```
-This specialist has learned patterns from: {directories}
-
-**[Domain] Patterns:**
-- File structure: {pattern}
-- Naming: {pattern}
-- Code style: {pattern}
-
-Apply these patterns when {doing agent's task}.
-```
-
-### 7. Create Maintenance Skill (auto)
-
-Auto-create companion directive skill for every specialist:
-
-**File**: `.claude/skills/{specialist-name}-maintenance/SKILL.md`
-
-```yaml
----
-name: {specialist-name}-maintenance
-description: Directives for maintaining {specialist-name} agent. Use when updating specialist knowledge.
----
-
-# {Specialist Name} Maintenance
-
-## Pattern Sources
-Directories: {list of directories}
-
-## Embedded Patterns
-{High-level summary of extracted patterns}
-
-## Update Process
-1. Re-run repomix on source directories
-2. Analyze for new/changed patterns
-3. Update agent system prompt
-```
-
-### 8. Offer Custom Skill Creation
-
-If custom skill selected in requirements:
-"This specialist needs a custom skill. Would you like to define it now?"
-- If yes: use skill-development skill to create it
-- If no: note as TODO in agent file
-
-## Agent File Structure
-
-### Complete Format
-
-```markdown
----
-name: agent-identifier
-description: |
-  [Role description with trigger keywords and responsibility scope].
-
-  <example>
-  user: "trigger1 | trigger2 | trigger3"
-  </example>
-model: inherit
-color: blue
-allowed-tools: Read, Glob, Grep
-skills: repomix-extraction, skill-name
----
-
-You are [agent role description]...
-
-**Your Core Responsibilities:**
-1. [Responsibility 1]
-2. [Responsibility 2]
-
-**Process:**
-[Step-by-step workflow]
-
-**Output Format:**
-[What to return]
-```
-
-## Frontmatter Fields
-
-### name (required)
-
-Agent identifier used for namespacing and invocation.
-
-**Format:** lowercase, numbers, hyphens only
-**Length:** 3-50 characters
-**Pattern:** Must start and end with alphanumeric
-
-**Good examples:**
-- `code-reviewer`
-- `test-generator`
-- `api-docs-writer`
-
-**Bad examples:**
-- `helper` (too generic)
-- `-agent-` (starts/ends with hyphen)
-- `my_agent` (underscores not allowed)
-- `ag` (too short)
-
-### description (required)
-
-Defines when Claude should trigger this agent. **Most critical field.**
-
-**Must include:**
-1. Concise description with trigger keywords and responsibility scope
-2. ONE `<example>` block with variant syntax for compression
-
-**Format (condensed pattern):**
-```yaml
-description: |
-  [Role description with trigger keywords].
-
-  <example>
-  user: "trigger1 | trigger2 | trigger3"
-  </example>
-```
-
-**Variant syntax:** Use `|` to separate trigger phrase variants within a single example. This compresses multiple triggering scenarios into minimal context.
-
-**Example:**
-```yaml
-description: |
-  Research specialist for web search, documentation lookup, external info gathering.
-
-  <example>
-  user: "Research [topic] | Find docs for [library] | What are best practices for [pattern]?"
-  </example>
-```
-
-**Do NOT include:**
-- Multiple `<example>` blocks (use variant syntax instead)
-- `<commentary>` blocks (description conveys reasoning)
-- Context or assistant response lines
-
-See `references/triggering-examples.md` for additional patterns.
-
-### model (required)
-
-Which model the agent should use.
-
-**Options:**
-- `inherit` - Use same model as parent (recommended)
-- `sonnet` - Claude Sonnet (balanced)
-- `opus` - Claude Opus (most capable)
-- `haiku` - Claude Haiku (fast)
-
-### color (required)
-
-Visual identifier for agent in UI.
-
-**Guidelines:**
-| Color | Use For |
-|-------|---------|
-| blue/cyan | Analysis, review |
-| green | Success-oriented tasks |
-| yellow | Caution, validation |
-| red | Critical, security |
-| magenta | Creative, generation |
-
-### allowed-tools (optional)
-
-Restrict agent to specific tools. **Principle of least privilege.**
-
-**Common tool sets:**
-- Read-only analysis: `Read, Grep, Glob`
-- Code generation: `Read, Write, Grep`
-- Testing: `Read, Bash, Grep`
-- Full access: Omit field to inherit all
-
-### skills (optional)
-
-Skills to auto-load when agent starts. Comma-separated list.
-
-## System Prompt Patterns
-
-Four patterns for agent system prompts. See `references/system-prompt-patterns.md` for templates.
-
-| Pattern | Use When |
-|---------|----------|
-| Analysis | Reviewing, auditing, researching |
-| Generation | Creating code, tests, docs |
-| Validation | Checking, verifying, linting |
-| Orchestration | Multi-step workflows |
-
-## AI-Assisted Agent Generation
-
-For complex agents, use this prompt template:
-
-```json
-{
-  "request": "[USER DESCRIPTION]",
-  "requirements": {
-    "core_intent": "Extract primary purpose",
-    "persona": "Define expert role for domain",
-    "system_prompt": {
-      "behavioral_boundaries": true,
-      "specific_methodologies": true,
-      "edge_case_handling": true,
-      "output_format": true
-    },
-    "identifier": "lowercase-hyphens, 3-50 chars",
-    "description": "trigger keywords + ONE condensed example block",
-    "example": "ONE <example> with variant syntax (trigger1 | trigger2 | trigger3)"
-  }
-}
-```
-
-## Key Principles
-
-- Specialists are READ-ONLY - they return information, main agent implements
-- Description must include WHEN to trigger (main agent uses this for dispatch)
-- Keep scope focused - better to have multiple narrow specialists than one broad one
-- Skills determine what knowledge/capabilities the specialist has access to
-- Use ONE condensed `<example>` block with variant syntax (`|`) for trigger phrases
-
-## Reference Files
-
-| File | Content |
-|------|---------|
-| `references/triggering-examples.md` | Example block anatomy and patterns |
-| `references/system-prompt-patterns.md` | Four agent pattern templates |
-| `examples/complete-agent-examples.md` | Full working agent examples |
-| `scripts/validate-agent.sh` | Validate agent file structure and content |
-
-## Agent-Skill Workflow Pattern
-
-When an agent has workflow skills, the agent profile should:
-- Define the prompt pattern that triggers the skill
-- Direct to "use" the skill (skills are automatically loaded into context)
-- NOT duplicate the workflow steps
-
-Example in agent file:
-```markdown
-## [Workflow Name]
-
-When main agent asks to [prompt pattern], use the [skill-name] skill.
-```
-
-**Why this works:**
-- Agent's `skills:` frontmatter lists skills to auto-load into context
-- Main agent triggers agent based on description examples
-- Agent body references skill by name - skill content already available
-- No duplication: skill owns workflow logic, agent owns triggering conditions
-
-This keeps agent files lean and workflow logic centralized in skills.

package/.claude/skills/specialist-builder/docs/agent-catalog.md

@@ -1,28 +0,0 @@
-# Agent Catalog Patterns
-
-Read this when: designing new subagents, choosing agent responsibilities, structuring multi-agent workflows.
-
-## Context Gathering Agents
-
-| Agent | Purpose |
-|-------|---------|
-| codebase-analyzer | Understanding existing code structure |
-| deep-research-agent | Comprehensive research before implementation |
-| git-diff-analyzer | Branch summaries, change impact |
-
-## Decision-Making Agents
-
-| Agent | Purpose |
-|-------|---------|
-| pair-programmer | Comparing solution approaches |
-| system-architect | Scalable design decisions |
-| root-cause-analyst | Debugging, finding source of issues |
-| code-reviewer | Unbiased review, quality gates |
-| refactoring-expert | Clean code transformations |
-
-## Agent Design Principles
-
-- Single responsibility per agent
-- Clear handoff boundaries
-- Context flows up (findings), instructions flow down (tasks)
-- Agents return findings, main agent implements