Refactor high-complexity React components in Dify frontend. Use when `pnpm analyze-component...
npx skills add skilluse/skilluse --skill "work"
Install specific skill from multi-skill repository
# Description
Work incrementally on Issues with Git worktree support for parallel agent workflows. Use when starting a coding session, continuing work, or running parallel agents.
# SKILL.md
name: work
description: Work incrementally on Issues with Git worktree support for parallel agent workflows. Use when starting a coding session, continuing work, or running parallel agents.
Work - Incremental Implementation
Make incremental progress on Issues, leaving clean state for next session.
Based on Anthropic's Coding Agent pattern.
Arguments
This skill accepts an optional issue path:
/work [issue-path]
Examples:
- /work - Continue work on current issue from Epic progress.md
- /work issues/20251227-task05-config-store - Work on specific issue
When to Use
- Starting a new coding session
- User says "continue", "work on issue", "next feature"
- Running AI agents on parallel issues (via worktrees)
Session Workflow
1. GET BEARINGS β pwd, git worktree list, cat epics/*/progress.md
β
2. READ ISSUE β Read issue.md and feature.json from issue directory
β
3. VERIFY BASELINE β Run build/tests, fix any broken state
β
4. IMPLEMENT β One feature at a time: implement β test β commit
β
5. END SESSION β Use /commit skill, update Epic progress.md
Directory Structure
project/
βββ epics/
β βββ cli/
β β βββ epic.md
β β βββ progress.md # Check here for current work
β β βββ references/
β βββ website/
β βββ epic.md
β βββ progress.md
β βββ references/
βββ issues/
βββ 20250115-task01-auth-setup/
β βββ issue.md # Requirements
β βββ feature.json # Acceptance criteria
βββ ...
Reading the Issue
Each issue directory contains:
- issue.md - Requirements, technical details, API design
- feature.json - Testable acceptance criteria with pass/fail status
# Read issue to understand the work
cat issues/ISSUE_ID/issue.md
cat issues/ISSUE_ID/feature.json
Core Loop: Implement One Feature
# 1. Implement the feature
# 2. Test end-to-end (browser/HTTP/CLI)
# 3. Update feature.json: "passes": true
# 4. Use /commit skill to commit changes
Rules
DO:
- Start every session with "Get Bearings"
- Read issue.md and feature.json before implementing
- Work on ONE feature at a time
- Test BEFORE marking passes: true
- Use /commit skill for commits (not manual git commit)
- Update Epic's progress.md at session end
DO NOT:
- Skip baseline verification
- Mark features passing without testing
- Leave half-implemented code
- Commit without using the commit skill
Launch Agent in New Terminal
When user says "work on [issue]" with a specific issue, open a new Ghostty terminal and run the agent there:
# Get current directory first
PROJ_DIR=$(pwd)
# Open Ghostty and run claude agent in that directory
open -a Ghostty --args -e zsh -lc "cd $PROJ_DIR && claude --dangerously-skip-permissions"
This allows:
- Running an autonomous agent in a separate terminal
- Continuing work in the current terminal
- Agent runs with full permissions for unattended operation
Parallel Development (Worktrees)
Use Git worktrees to run agents on separate issues in parallel:
# Create worktree for agent
git worktree add ../project-task02 -b task02
cd ../project-task02
claude --print "Complete task02 per issue.md"
# Continue your work in main repo
cd ../project
See: worktree-commands.md | parallel-agents.md
Handling Issues
| Situation | Action |
|---|---|
| Broken baseline | Fix FIRST, then new work |
| Blocked by dependency | Document, switch issue or end session |
| Feature fails testing | Keep passes: false, fix and re-test |
Progress File Format
Update the Epic's progress.md:
#### [YYYY-MM-DD HH:MM] Session #N
- Working on: task02-auth
- Completed: F001, F002
- In progress: F003
- Commits: abc123, def456
- Notes:
- Brief description of what was done
- Any important decisions or changes
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.