Agent skill
project_workflow
Create or update project workflow documentation in AGENTS.md (for opencode/codex) or CLAUDE.md (for claude). Use when: user asks to create/update a workflow, document build/test/debug steps, or update project documentation. The skill will check memory and repo to extract relevant commands, then write sections to the appropriate file. NOT for: one-off commands, reading existing docs only, or when the workflow sections are already complete.
Install this agent skill to your Project
npx add-skill https://github.com/deciding/handclaw/tree/main/skills/project_workflow
Metadata
Additional technical details for this skill
- openclaw
-
{ "emoji": "\ud83d\udccb", "requires": [] }
SKILL.md
Project Workflow Skill
Create or update project workflow documentation in the repo's AGENTS.md (for opencode/codex) or CLAUDE.md (for claude).
When to Use
✅ USE this skill when:
- User asks to "create a workflow" or "document the workflow"
- User asks to add "build", "test", "debug" steps to AGENTS.md/CLAUDE.md
- User asks to document how to build, test, or debug the project
- Setting up a new project and need to document development workflow
- User asks to add "command cheatsheet", "workflow process", or "mistake list"
When NOT to Use
❌ DON'T use this skill when:
- User only wants to run a single build/test command
- User is asking about a specific error or bug (use debugging tools)
- AGENTS.md/CLAUDE.md already has complete workflow sections and user just wants to read them
- User wants to commit/push workflow docs (that's a git operation)
Process
Step 1: Detect Agent Type
Determine which agent is being used:
- opencode or codex → use
AGENTS.md - claude → use
CLAUDE.md
Step 2: Check Memory
Look for relevant context in the agent's memory:
- Check for any remembered build commands, test commands, or debug tips
- Look for package manager info (pnpm, npm, bun, yarn)
- Check for any stored configuration or environment setup notes
- Check for previous mistakes and solutions
Step 3: Check the Repo
Examine the repository to find workflow-relevant information:
- package.json: Look for scripts (build, test, dev, lint, typecheck)
- README.md: Check for build/test instructions
- AGENTS.md or CLAUDE.md: Check if workflow sections already exist
- Makefile or justfile: Check for build recipes
- Dockerfile: Check for build steps if applicable
Step 4: Create ~/.handclaw Directory (if needed)
For long or reusable content, create a .handclaw directory in the user's home directory:
~/.handclaw/WORKFLOW.md- Full workflow process documentation~/.handclaw/TIPS.md- Debug tips and common mistakes~/.handclaw/COMMANDS.md- Extended command reference
Step 5: Write Sections to AGENTS.md/CLAUDE.md
Create or update the following sections at the END of the appropriate file:
1. Command Cheatsheet
## Command Cheatsheet
- Install deps: `<package-manager> install`
- Build: `<package-manager> build` if failed check solutions in @~/.handclaw/COMMANDS.md
- Dev: `<package-manager> dev`
- Test: `<package-manager> test`
- Lint: `<package-manager> lint`
- Typecheck: `<package-manager> typecheck`
If content is short, keep inline. If long, reference external file.
2. Workflow
## Workflow
1. **Plan**: Understand the task, analyze requirements, check existing code
2. **Todo**: Create a todo list with `TodoWrite` for multi-step tasks
3. **Change Code**: Implement changes step by step
4. **Review**: Review changes, run lint/typecheck
5. **Build/Test**: If successful → build → test → deploy
- If problems → go back to step 2 or 3
6. **Commit** (only when explicitly asked)
For detailed workflow see @~/.handclaw/WORKFLOW.md
3. Mistake List
## Mistake List
- **[Error/Issue]**: [Solution/Tip]
- **Module Z failed**: add a debug=True parameter in the function xyz calling
- **Module X reports error Y**: Check Z as described in @~/.handclaw/TIPS.md
Section Update Logic
- If a section already exists, update it with current information
- If a section doesn't exist, create it
- Preserve existing content that isn't part of these sections
- Create the file (AGENTS.md or CLAUDE.md) if it doesn't exist
- Place all new sections at the END of the file
Output Format
Write sections like:
---
## Command Cheatsheet
- Install deps: `pnpm install`
- Build: `pnpm build`
- Dev: `pnpm dev`
- Test: `pnpm test`
- Lint: `pnpm lint`
- Typecheck: `pnpm typecheck`
## Workflow
1. **Plan**: Understand the task, analyze requirements, check existing code
2. **Todo**: Create a todo list with `TodoWrite` for multi-step tasks
3. **Change Code**: Implement changes step by step
4. **Review**: Review changes, run lint/typecheck
5. **Build/Test**: If successful → build → test → deploy
- If problems → go back to step 2 or 3
6. **Commit** (only when explicitly asked)
## Mistake List
- **pnpm install fails**: Try deleting node_modules and pnpm-lock.yaml first
- **TypeScript errors**: Run `pnpm typecheck` to see all errors
- **Build fails**: Check if all dependencies are installed
For longer content, use references:
## Command Cheatsheet
If case1 encountered, See @~/.handclaw/COMMANDS.md for the `solutionToCase1` command.
Notes
- Always use the actual commands found in the memory/repo, don't guess
- Prioritize memory over repo
- Add project-specific debug tips if discovered
- Keep sections concise but actionable
- Use references (@~/.handclaw/...) for long or reusable content
- The keyword format
@~/.handclaw/FILE.mdallows quick lookup
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
deep-research
Use this skill instead of WebSearch for ANY question requiring web research. Trigger on queries like "what is X", "explain X", "compare X and Y", "research X", or before content generation tasks. Provides systematic multi-angle research methodology instead of single superficial searches. Use this proactively when the user's question needs online information.
verl-rl-training
Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.
openrlhf-training
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
gguf-quantization
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
Claude Code Guide
Master guide for using Claude Code effectively. Includes configuration templates, prompting strategies "Thinking" keywords, debugging techniques, and best practices for interacting with the agent.
qdrant-vector-search
High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.
Didn't find tool you were looking for?