-
Notifications
You must be signed in to change notification settings - Fork 2
FAQ
Frequently asked questions about ai-coding-ok, grouped by topic.
No. Skills are a native Claude Code feature available in the free tier. Copilot users only need a Copilot subscription. ai-coding-ok itself is completely free (MIT license).
Yes, with varying levels of integration:
| Tool | How | PDCA Strength |
|---|---|---|
| Cursor |
.cursor/rules/ai-coding-ok.mdc (alwaysApply) |
🛡️🛡️🛡️ |
| Continue | Reference AGENTS.md in rules config |
🛡️🛡️ |
| Cline | Reference AGENTS.md in rules config |
🛡️🛡️ |
The memory files (.github/agent/memory/*.md) are plain Markdown — any AI tool can read them.
Use the Python install script:
python install.py --copilot --target C:\your-projectOr use Git Bash to run install.sh.
No. The installer detects conflicts and offers three choices:
- Overwrite (risky)
- Copy only missing files (safe, recommended)
- Abort
It never silently overwrites existing files.
Yes. It adds files to .github/agent/ and creates/updates AGENTS.md and CLAUDE.md. It doesn't modify your source code or build configuration.
The three memory files combined are typically <10KB — negligible in the AI's context window. The cost of NOT reading memory (wrong decisions, breaking existing features) is far higher.
No. The convention is to keep the last 30 entries. Older entries should be archived to docs/task-history-archive-YYYY-QN.md. The AI's system prompt instructs it to auto-archive.
Usually no. The AI updates task-history.md automatically after every task (Act phase). You should:
- Occasionally review AI-written entries for accuracy
- Manually write ADRs for major architecture decisions
- Monthly scan
project-memory.mdfor stale information
In Claude Code, the Stop hook (v3.1.0+) blocks session end if memory wasn't updated. For Copilot/Cursor, the mandatory ## Memory Updates output section (v3.0.1+) ensures the AI can't skip it.
If it still happens, manually remind:
Run the Act phase: update task-history.md with this task's summary.
Yes. ai-coding-ok has full bilingual support (English + Chinese) since v3.0.0. The AI auto-detects your language during install.
Claude Code: upgrade ai-coding-ok in any project.
Copilot/Cursor: Paste scripts/upgrade-prompt.md into your AI tool's chat.
The upgrade preserves all your project customizations. See Upgrade ai-coding-ok.
No. The upgrade system diffs at the section level and only applies framework changes. Your project name, tech stack, architecture diagrams, custom ADRs, and constraints are all preserved.
head -1 AGENTS.md
# → <!-- ai-coding-ok: v3.1.0 -->v3.0.0 introduced plugin packaging and bilingual templates. v3.1.0 added the Five-Layer Defense System with Claude Code hooks — the most significant reliability improvement since v2.0.
See Version Management for the full history.
No. They solve different problems:
- superpowers = deep thinking within a session (brainstorming, research, multi-agent)
- ai-coding-ok = cross-session memory (remembering context, decisions, task history)
They compose cleanly. See Combo with superpowers.
Yes. ai-coding-ok works independently — you lose deep research/brainstorming capabilities. superpowers works independently — you lose cross-session memory.
If you can only learn one, start with ai-coding-ok (the memory problem is more urgent).
ai-coding-ok v2.0+ has built-in superpowers compatibility:
- AGENTS.md hooks into brainstorming's Step 1
- Plans generated by superpowers auto-include a memory update task
- ai-coding-ok ensures Act runs after superpowers finishes
- Fork the ai-coding-ok repo
- Edit files in
templates/en/andtemplates/zh/(both must be synced) - If behavior changes, also edit
skills/ai-coding-ok/SKILL.md - Submit a PR describing what you changed and why
Templates are defaults — you're expected to customize them. If the default is genuinely wrong for a common project type, open an issue to discuss adding a variant.
Open an issue on GitHub. Include:
- Your ai-coding-ok version (
head -1 AGENTS.md) - Your AI tool (Claude Code / Copilot / Cursor)
- What you expected vs what happened
- Any error messages
# Remove installed files
rm AGENTS.md CLAUDE.md
rm -rf .github/copilot-instructions.md .github/project-metadata.yml \
.github/PULL_REQUEST_TEMPLATE.md .github/ISSUE_TEMPLATE \
.github/workflows/ci.yml .github/workflows/memory-check.yml \
.github/agent
rm -rf .cursor/rules/ai-coding-ok.mdc
# If .github/workflows/ and .github/ are now empty, remove them tooOr git reset --hard HEAD~1 if you committed the install and haven't made other changes.
Remove the .github/agent/memory/ directory. The rest of the rules (coding standards, workflows) still apply, but the PDCA memory loop won't trigger.
To re-enable: re-install or restore the directory from git.
No. It's entirely local — Markdown files in your project and shell scripts that run locally. No telemetry, no API calls, no external services.
Never. Memory files are committed to git. Use environment variables or a .env file (gitignored) for secrets.
Yes. MIT license allows commercial use, private repos, and modifications. No attribution required (though appreciated).
- Troubleshooting — common issues and fixes
- GitHub Issues — ask the community
- Daily Workflow — practical usage guide
🧠 ai-coding-ok — AI 编程的 PDCA 记忆闭环。
GitHub · Issues · MIT License