Turn your AI coding agent into a reliable engineer.
Super Pi is a Pi-native engineering workflow layer: it adds stage discipline, durable artifacts, TDD gates, checkpoints, review, and learning loops on top of your coding agent. It is not a general-purpose multi-agent executor; subagents are optional helpers for bounded leaf tasks.
Install, describe what you want to build, then keep saying "continue." Super Pi drives the full loop:
think → plan → build → review → compound learnings.
pi install npm:@leing2021/super-pi- Five-step loop — brainstorm → plan → work → review → learn, with automatic skill routing
- Checkpoint resume — interrupted? Resume from the exact unit you left off
- TDD enforcement — every unit follows RED → GREEN → REFACTOR with hard gates
- Controlled subagents — inline execution by default; subagents are limited to bounded, non-interactive leaf tasks
- Evidence-first review — auto-assigned reviewers across five axes, autofix loop
- Knowledge compounding — solved problems become searchable solution artifacts
- Token-efficient — ~2,600 tokens new-conversation overhead; progressive loading
pi install npm:@leing2021/super-piThen in Pi:
You: I want to build a CLI tool that helps indie devs find early users
→ 01-brainstorm: structured discovery → requirements artifact
→ 02-plan: TDD-gated implementation units → plan artifact
→ 03-work: inline-first execution, bounded subagents, checkpoint resume
→ 04-review: five-axis findings, autofix loop
→ 05-learn: knowledge compounding
You: continue
→ Next skill recommended via /skill:06-next
Resume after interruption:
You: /skill:03-work docs/plans/plan.md
→ Loads checkpoint, skips completed units, resumes from breakpoint
01-brainstorm → 02-plan → 03-work → 04-review → 05-learn
think plan build review learn
| Skill | What it does | Core tool |
|---|---|---|
| 01-brainstorm | Structured multi-round discovery, domain vocabulary persistence | brainstorm_dialog |
| 02-plan | TDD-gated implementation units, optional CEO Review | plan_diff |
| 03-work | Inline-first execution, bounded subagents, checkpoint resume, strict TDD, stop-the-line | ce_subagent, ce_parallel_subagent |
| 04-review | Auto-assigned reviewers, five-axis findings, autofix loop | review_router |
| 05-learn | Pattern extraction → searchable solution artifacts | pattern_extractor |
| 06-next | Next-step recommendation + workflow status | workflow_state |
| 07-worktree | Isolated git worktree development | worktree_manager |
| 08-help | Phase 1 skill explainer | — |
Super Pi ships CE-specific tools named ce_subagent and ce_parallel_subagent. They are intentionally namespaced so they can coexist with the compatible third-party pi-subagents extension without tool-name collisions.
These tools are helpers, not the core workflow. Use inline execution by default. Use CE subagents only for bounded, non-interactive, easily verifiable leaf tasks. Do not use them to invoke pipeline-stage skills (01-brainstorm through 05-learn); run those stages directly with /skill:<stage>.
Configure in .pi/settings.json:
{
"modelStrategy": {
"01-brainstorm": "anthropic/claude-sonnet-4-20250514",
"02-plan": "anthropic/claude-opus-4-20250115"
},
"thinkingStrategy": {
"01-brainstorm": "high",
"02-plan": "high",
"03-work": "medium"
}
}Model and thinking level switch automatically — no manual /model needed.
80% planning and review, 20% execution.
The goal is not to make AI write code faster. The goal is to make AI think before writing, review after writing, and compound what it learns.
Super Pi is not a fork or wrapper. It extracts useful methods from the projects below and rebuilds them with Pi-native skills, tools, artifacts, checkpoints, and handoffs.
| Project | What Super Pi adopted |
|---|---|
| addyosmani/agent-skills | "Use when" skill trigger conditions, source-driven verification, stop-the-line hard gate, anti-rationalization, and the five-axis review baseline. Adopted as embedded micro-patterns only — no new skills, tools, commands, or agents. |
| everything-claude-code | Parallel subagent orchestration, checkpoint resume, continuous learning loops, and token-conscious agent workflow design. |
| humanlayer/12-factor-agents | Context window ownership, compacting resolved errors, retry caps, and pre-fetching obvious prerequisites. Adopted as lightweight context hygiene rules inside the existing Phase 1 pipeline. |
| superpowers | Strict TDD gates, design checklists, review discipline, and the idea that agents need hard gates instead of gentle suggestions. |
| compound-engineering-plugin | The five-step think → plan → build → review → learn loop and the knowledge-compounding backbone. |
| gstack | YC-style forcing questions, CEO Review cognitive frameworks, browser QA patterns, failure maps, and evidence-first validation. |
| mattpocock/skills | Context glossary (CONTEXT.md) for cross-session term persistence, lightweight ADR with three-condition threshold, and feedback-loop-first debug discipline. Adopted as reference templates embedded into existing skills — no new skills or tools. |
When an unexpected failure occurs during 03-work:
- STOP adding features
- PRESERVE evidence
- DIAGNOSE root cause — build a feedback loop first, then reproduce → hypothesise → instrument → fix
- FIX the root cause, not the symptom
- GUARD with a regression test
- RESUME only after verification passes
Anti-rationalization: do not rationalize, downgrade, or explain away failures. Stop and report with evidence.
When implementation depends on a framework/library API, version-specific behavior, or a recommended pattern: verify against official documentation before implementing. Pure logic, renaming, or in-project pattern reuse does not require external citation.
All reviewers evaluate changes across: correctness, readability, architecture, security, performance.
New conversation overhead: ~2,600 tokens (1.3% of 200K context).
| Component | Tokens |
|---|---|
| 8 skill registrations | ~490 |
| System prompt | ~1,400 |
| Skill inlining (per invocation) | ~500-800 |
Progressive loading: only needed skills loaded on-demand.
See docs/token-cost-evaluation.md for detailed per-skill breakdown and measurement methodology.
your-project/
├── docs/
│ ├── brainstorms/ # Requirements
│ ├── plans/ # Execution plans
│ ├── adr/ # Architecture decisions (lazy)
│ └── solutions/ # Knowledge cards
└── .context/
└── compound-engineering/
├── checkpoints/ # Breakpoint files
├── handoffs/ # Cross-stage context
└── history/ # Execution history
Commit everything to git — these files are the project's traceable memory.
| Component | Count |
|---|---|
| Skills | 8 |
| Tools | 19 |
| Rules | 79 |
| TypeScript lines | ~4,400 |
| Tests | 209 (786 assertions) |
Rules in rules/ cover 11 common topics + language-specific sets (TypeScript, Rust, Go, Python, Java, Kotlin, C++, C#, Dart, Swift, Perl, PHP). Project-level overrides take priority.
| Command | Description |
|---|---|
bun test |
Run all tests |
npm publish --dry-run |
Preview package contents |
See CHANGELOG.md for full version history.
- npm: https://www.npmjs.com/package/@leing2021/super-pi
- GitHub: https://github.com/leing2021/super-pi
- License: MIT