Claude-Code-Workflow/.claude/commands/memory/skill-memory.md

name: skill-memory description: 4-phase autonomous orchestrator: check docs → /memory:docs planning → /workflow:execute → generate SKILL.md with progressive loading index (skips phases 2-3 if docs exist) argument-hint: "[path] [--tool <gemini|qwen|codex>] [--regenerate] [--mode <full|partial>] [--cli-execute]" allowed-tools: SlashCommand(), TodoWrite(), Bash(), Read(), Write(*)

Memory SKILL Package Generator

Orchestrator Role

Pure Orchestrator: Execute documentation generation workflow, then generate SKILL.md index. Does NOT create task JSON files.

Auto-Continue Workflow: This command runs fully autonomously once triggered. Each phase completes and automatically triggers the next phase without user interaction.

Execution Paths:

• Full Path: All 4 phases (no existing docs OR --regenerate specified)
• Skip Path: Phase 1 → Phase 4 (existing docs found AND no --regenerate flag)
• Phase 4 Always Executes: SKILL.md index is never skipped, always generated or updated

Core Rules

1. Start Immediately: First action is TodoWrite initialization, second action is Phase 1 execution
2. No Task JSON: This command does not create task JSON files - delegates to /memory:docs
3. Parse Every Output: Extract required data from each command output (session_id, task_count, file paths)
4. Auto-Continue: After completing each phase, update TodoWrite and immediately execute next phase
5. Track Progress: Update TodoWrite after EVERY phase completion before starting next phase
6. Direct Generation: Phase 4 directly generates SKILL.md using Write tool
7. No Manual Steps: User should never be prompted for decisions between phases

---

4-Phase Execution

Phase 1: Prepare Arguments

Goal: Parse command arguments and check existing documentation

Step 1: Get Target Path and Project Name

# Get current directory (or use provided path)
bash(pwd)

# Get project name from directory
bash(basename "$(pwd)")

# Get project root
bash(git rev-parse --show-toplevel 2>/dev/null || pwd)

Output:

• target_path: /d/my_project
• project_name: my_project
• project_root: /d/my_project

Step 2: Set Default Parameters

# Default values (use these unless user specifies otherwise):
# - tool: "gemini"
# - mode: "full"
# - regenerate: false (no --regenerate flag)
# - cli_execute: false (no --cli-execute flag)

Step 3: Check Existing Documentation

# Check if docs directory exists
bash(test -d .workflow/docs/my_project && echo "exists" || echo "not_exists")

# Count existing documentation files
bash(find .workflow/docs/my_project -name "*.md" 2>/dev/null | wc -l || echo 0)

Output:

• docs_exists: exists or not_exists
• existing_docs: 5 (or 0 if no docs)

Step 4: Determine Execution Path

Decision Logic:

if (existing_docs > 0 && !regenerate_flag) {
  // Documentation exists and no regenerate flag
  SKIP_DOCS_GENERATION = true
  message = "Documentation already exists, skipping Phase 2 and Phase 3. Use --regenerate to force regeneration."
} else if (regenerate_flag) {
  // Force regeneration: delete existing docs
  bash(rm -rf .workflow/docs/my_project 2>/dev/null || true)
  SKIP_DOCS_GENERATION = false
  message = "Regenerating documentation from scratch."
} else {
  // No existing docs
  SKIP_DOCS_GENERATION = false
  message = "No existing documentation found, generating new documentation."
}

Summary Variables:

• PROJECT_NAME: my_project
• TARGET_PATH: /d/my_project
• DOCS_PATH: .workflow/docs/my_project
• TOOL: gemini (default) or user-specified
• MODE: full (default) or user-specified
• CLI_EXECUTE: false (default) or true if --cli-execute flag
• REGENERATE: false (default) or true if --regenerate flag
• EXISTING_DOCS: Count of existing documentation files
• SKIP_DOCS_GENERATION: true if skipping Phase 2/3, false otherwise

Completion & TodoWrite:

• If SKIP_DOCS_GENERATION = true: Mark phase 1 completed, phase 2&3 completed (skipped), phase 4 in_progress
• If SKIP_DOCS_GENERATION = false: Mark phase 1 completed, phase 2 in_progress

Next Action:

• If skipping: Display skip message → Jump to Phase 4 (SKILL.md generation)
• If not skipping: Display preparation results → Continue to Phase 2 (documentation planning)

---

Phase 2: Call /memory:docs

Skip Condition: This phase is skipped if SKIP_DOCS_GENERATION = true (documentation already exists without --regenerate flag)

Goal: Trigger documentation generation workflow

Command:

SlashCommand(command="/memory:docs [targetPath] --tool [tool] --mode [mode] [--cli-execute]")

Example:

/memory:docs /d/my_app --tool gemini --mode full
/memory:docs /d/my_app --tool gemini --mode full --cli-execute

Note: The --regenerate flag is handled in Phase 1 by deleting existing documentation. This command always calls /memory:docs without the regenerate flag, relying on docs.md's built-in update detection.

Parse Output:

• Extract session ID: WFS-docs-[timestamp] (store as docsSessionId)
• Extract task count (store as taskCount)

Completion Criteria:

• /memory:docs command executed successfully
• Session ID extracted and stored
• Task count retrieved
• Task files created in .workflow/[docsSessionId]/.task/
• workflow-session.json exists

TodoWrite: Mark phase 2 completed, phase 3 in_progress

Next Action: Display docs planning results (session ID, task count) → Auto-continue to Phase 3

---

Phase 3: Execute Documentation Generation

Skip Condition: This phase is skipped if SKIP_DOCS_GENERATION = true (documentation already exists without --regenerate flag)

Goal: Execute documentation generation tasks

Command:

SlashCommand(command="/workflow:execute")

Note: /workflow:execute automatically discovers active session from Phase 2

Completion Criteria:

• /workflow:execute command executed successfully
• Documentation files generated in .workflow/docs/[projectName]/
• All tasks marked as completed in session
• At minimum: module documentation files exist (API.md and/or README.md)
• For full mode: Project README, ARCHITECTURE, EXAMPLES files generated

TodoWrite: Mark phase 3 completed, phase 4 in_progress

Next Action: Display execution results (file count, module count) → Auto-continue to Phase 4

---

Phase 4: Generate SKILL.md Index

Note: This phase is NEVER skipped - it always executes to generate or update the SKILL index.

Step 1: Read Key Files (Use Read tool)

• .workflow/docs/{project_name}/README.md (required)
• .workflow/docs/{project_name}/ARCHITECTURE.md (optional)

Step 2: Discover Structure

bash(find .workflow/docs/{project_name} -name "*.md" | sed 's|.workflow/docs/{project_name}/||' | awk -F'/' '{if(NF>=2) print $1"/"$2}' | sort -u)

Step 3: Generate Intelligent Description

Extract from README + structure: Function (capabilities), Modules (names), Keywords (API/CLI/auth/etc.)

Format: {Project} {core capabilities} (located at {project_path}). Load this SKILL when analyzing, modifying, or learning about {domain_description} or files under this path, especially when no relevant context exists in memory.

Key Elements:

• Path Reference: Use TARGET_PATH from Phase 1 for precise location identification
• Domain Description: Extract human-readable domain/feature area from README (e.g., "workflow management", "thermal modeling")
• Trigger Optimization: Include project path, emphasize "especially when no relevant context exists in memory"
• Action Coverage: analyzing (分析), modifying (修改), learning (了解)

Example: "Workflow orchestration system with CLI tools and documentation generation (located at /d/Claude_dms3). Load this SKILL when analyzing, modifying, or learning about workflow management or files under this path, especially when no relevant context exists in memory."

Step 4: Write SKILL.md (Use Write tool)

bash(mkdir -p .claude/skills/{project_name})

.claude/skills/{project_name}/SKILL.md:

---
name: {project_name}
description: {intelligent description from Step 3}
version: 1.0.0
---
# {Project Name} SKILL Package

## Documentation: `../../../.workflow/docs/{project_name}/`

## Progressive Loading
### Level 0: Quick Start (~2K)
- [README](../../../.workflow/docs/{project_name}/README.md)
### Level 1: Core Modules (~8K)
{Module READMEs}
### Level 2: Complete (~25K)
All modules + [Architecture](../../../.workflow/docs/{project_name}/ARCHITECTURE.md)
### Level 3: Deep Dive (~40K)
Everything + [Examples](../../../.workflow/docs/{project_name}/EXAMPLES.md)

Completion Criteria:

• SKILL.md file created at .claude/skills/{project_name}/SKILL.md
• Intelligent description generated from documentation
• Progressive loading levels (0-3) properly structured
• Module index includes all documented modules
• All file references use relative paths

TodoWrite: Mark phase 4 completed

Final Action: Report completion summary to user

Return to User:

SKILL Package Generation Complete

Project: {project_name}
Documentation: .workflow/docs/{project_name}/ ({doc_count} files)
SKILL Index: .claude/skills/{project_name}/SKILL.md

Generated:
- {task_count} documentation tasks completed
- SKILL.md with progressive loading (4 levels)
- Module index with {module_count} modules

Usage:
- Load Level 0: Quick project overview (~2K tokens)
- Load Level 1: Core modules (~8K tokens)
- Load Level 2: Complete docs (~25K tokens)
- Load Level 3: Everything (~40K tokens)

---

Implementation Details

Critical Rules

1. No User Prompts Between Phases: Never ask user questions or wait for input between phases
2. Immediate Phase Transition: After TodoWrite update, immediately execute next phase command
3. Status-Driven Execution: Check TodoList status after each phase:
   • If next task is "pending" → Mark it "in_progress" and execute
   • If all tasks are "completed" → Report final summary
4. Phase Completion Pattern:

   Phase N completes → Update TodoWrite (N=completed, N+1=in_progress) → Execute Phase N+1
   

TodoWrite Patterns

Initialization (Before Phase 1)

FIRST ACTION: Create TodoList with all 4 phases

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "in_progress", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "pending", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "pending", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "pending", "activeForm": "Generating SKILL.md"}
]})

SECOND ACTION: Execute Phase 1 immediately

Full Path (SKIP_DOCS_GENERATION = false)

After Phase 1:

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "in_progress", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "pending", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "pending", "activeForm": "Generating SKILL.md"}
]})
// Auto-continue to Phase 2

After Phase 2:

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "completed", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "in_progress", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "pending", "activeForm": "Generating SKILL.md"}
]})
// Auto-continue to Phase 3

After Phase 3:

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "completed", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "completed", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "in_progress", "activeForm": "Generating SKILL.md"}
]})
// Auto-continue to Phase 4

After Phase 4:

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "completed", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "completed", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "completed", "activeForm": "Generating SKILL.md"}
]})
// Report completion summary to user

Skip Path (SKIP_DOCS_GENERATION = true)

After Phase 1 (detects existing docs, skips Phase 2 & 3):

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "completed", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "completed", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "in_progress", "activeForm": "Generating SKILL.md"}
]})
// Display skip message: "Documentation already exists, skipping Phase 2 and Phase 3. Use --regenerate to force regeneration."
// Jump directly to Phase 4

After Phase 4:

TodoWrite({todos: [
  {"content": "Parse arguments and prepare", "status": "completed", "activeForm": "Parsing arguments"},
  {"content": "Call /memory:docs to plan documentation", "status": "completed", "activeForm": "Calling /memory:docs"},
  {"content": "Execute documentation generation", "status": "completed", "activeForm": "Executing documentation"},
  {"content": "Generate SKILL.md index", "status": "completed", "activeForm": "Generating SKILL.md"}
]})
// Report completion summary to user

Execution Flow Diagrams

Full Path Flow

User triggers command
  ↓
[TodoWrite] Initialize 4 phases (Phase 1 = in_progress)
  ↓
[Execute] Phase 1: Parse arguments
  ↓
[TodoWrite] Phase 1 = completed, Phase 2 = in_progress
  ↓
[Execute] Phase 2: Call /memory:docs
  ↓
[TodoWrite] Phase 2 = completed, Phase 3 = in_progress
  ↓
[Execute] Phase 3: Call /workflow:execute
  ↓
[TodoWrite] Phase 3 = completed, Phase 4 = in_progress
  ↓
[Execute] Phase 4: Generate SKILL.md
  ↓
[TodoWrite] Phase 4 = completed
  ↓
[Report] Display completion summary

Skip Path Flow

User triggers command
  ↓
[TodoWrite] Initialize 4 phases (Phase 1 = in_progress)
  ↓
[Execute] Phase 1: Parse arguments, detect existing docs
  ↓
[TodoWrite] Phase 1 = completed, Phase 2&3 = completed (skipped), Phase 4 = in_progress
  ↓
[Display] Skip message: "Documentation already exists, skipping Phase 2 and Phase 3"
  ↓
[Execute] Phase 4: Generate SKILL.md (always runs)
  ↓
[TodoWrite] Phase 4 = completed
  ↓
[Report] Display completion summary

Error Handling

• If any phase fails, mark it as "in_progress" (not completed)
• Report error details to user
• Do NOT auto-continue to next phase on failure

---

Parameters

/memory:skill-memory [path] [--tool <gemini|qwen|codex>] [--regenerate] [--mode <full|partial>] [--cli-execute]

• path: Target directory (default: current directory)
• --tool: CLI tool for documentation (default: gemini)
  • gemini: Comprehensive documentation
  • qwen: Architecture analysis
  • codex: Implementation validation
• --regenerate: Force regenerate all documentation
  • When enabled: Deletes existing .workflow/docs/{project_name}/ before regeneration
  • Ensures fresh documentation from source code
• --mode: Documentation mode (default: full)
  • full: Complete docs (modules + README + ARCHITECTURE + EXAMPLES)
  • partial: Module docs only
• --cli-execute: Enable CLI-based documentation generation (optional)
  • When enabled: CLI generates docs directly in implementation_approach
  • When disabled (default): Agent generates documentation content

---

Examples

Example 1: Generate SKILL Package (Default)

/memory:skill-memory

Workflow:

1. Phase 1: Detects current directory, checks existing docs
2. Phase 2: Calls /memory:docs . --tool gemini --mode full (Agent Mode)
3. Phase 3: Executes documentation generation via /workflow:execute
4. Phase 4: Generates SKILL.md at .claude/skills/{project_name}/SKILL.md

Example 2: Regenerate with Qwen

/memory:skill-memory /d/my_app --tool qwen --regenerate

Workflow:

1. Phase 1: Parses target path, detects regenerate flag, deletes existing docs
2. Phase 2: Calls /memory:docs /d/my_app --tool qwen --mode full
3. Phase 3: Executes documentation regeneration
4. Phase 4: Generates updated SKILL.md

Example 3: Partial Mode (Modules Only)

/memory:skill-memory --mode partial

Workflow:

1. Phase 1: Detects partial mode
2. Phase 2: Calls /memory:docs . --tool gemini --mode partial (Agent Mode)
3. Phase 3: Executes module documentation only
4. Phase 4: Generates SKILL.md with module-only index

Example 4: CLI Execute Mode

/memory:skill-memory --cli-execute

Workflow:

1. Phase 1: Detects CLI execute mode
2. Phase 2: Calls /memory:docs . --tool gemini --mode full --cli-execute (CLI Mode)
3. Phase 3: Executes CLI-based documentation generation
4. Phase 4: Generates SKILL.md at .claude/skills/{project_name}/SKILL.md

Example 5: Skip Path (Existing Docs)

/memory:skill-memory

Scenario: Documentation already exists in .workflow/docs/{project_name}/

Workflow:

1. Phase 1: Detects existing docs (5 files), sets SKIP_DOCS_GENERATION = true
2. Display: "Documentation already exists, skipping Phase 2 and Phase 3. Use --regenerate to force regeneration."
3. Phase 4: Generates or updates SKILL.md index only (~5-10x faster)

---

Architecture

skill-memory (orchestrator)
  ├─ Phase 1: Prepare (bash commands, skip decision)
  ├─ Phase 2: /memory:docs (task planning, skippable)
  ├─ Phase 3: /workflow:execute (task execution, skippable)
  └─ Phase 4: Write SKILL.md (direct file generation, always runs)

No task JSON created by this command
All documentation tasks managed by /memory:docs
Smart skip logic: 5-10x faster when docs exist