/pi

description: PI Wisdom-in-Action Engine v23 — top mandates, scene router, delivery quality gates, anti-patterns, battle momentum, Loop/Auto interaction plus Wenyan output. Invoke for structured execution with concise evidence.

Unless the first argument is `visualize`, invoke the pi skill and follow it exactly as presented to you.

Project Context (Auto-Injected)

> The following is auto-injected at invocation to assist 启动三查·查境. Use it to understand the current project state.

**Current branch**: !`git branch --show-current 2>/dev/null || echo '(not a git repo)'`

**Recent commits**: !`git log --oneline -5 2>/dev/null || echo '(no git history)'`

**Working state**: !`git diff --stat HEAD 2>/dev/null | tail -5 || echo '(clean or not a git repo)'`

**Project files** (top-level): !`ls -1 2>/dev/null | head -15 || echo '(empty)'`

Argument Routing

If the user provided arguments after `/pi`, parse them as follows:

| Argument | Effect | |----------|--------| | `visualize` | Launch the PI decision visualizer instead of the normal PI skill flow | | `loop` / `循环` / `接续` | Activate **Loop mode** (§8.2) — keep iterative sessions alive with concrete follow-up questions | | `auto` | Activate **Auto mode** (§8.2) — default autonomous interaction | | `deep` / `深度` | Force **Deep mode** (§8.2) — full investigation/review/debug depth | | `wenyan` / `文言` / `古文` | Less talk, more action; compressed Wenyan output; keep code/commands literal | | `dev` / `code` / `编程` / `开发` | Force **Coding & Development** scene (§1.3) | | Scene keyword (fleet/测试/产品/运营/创意/交互/调试/协作/陪伴) | Force-activate that scene (§1.3), skip auto-routing | | Any other text | Treat as the user's task description, proceed with auto scene routing |

Multiple arguments combine: `/pi loop 编程 文言` = Loop interaction + Coding scene + Wenyan output. Example: `/pi loop 编程 文言 深度` = Loop interaction + Coding scene + Wenyan output + Deep mode.

If no arguments provided, default to Auto mode with auto scene routing.

Special Route: `/pi visualize`

If the first argument is `visualize`, **do not** activate the generic PI scene router. Instead, treat the remaining arguments as visualizer CLI flags and execute the local visualizer flow:

1. Locate the visualizer tool:

• If `./visualize/package.json` exists in current path, use `./visualize`.
• Else if `~/.pi/visualize.sh` exists, use that directly.
• Else if `~/.pi/visualize/visualize` exists and contains `package.json`, use that.
• Else, fail and tell the user to run `curl -fsSL https://raw.githubusercontent.com/share-skills/pi/main/scripts/setup-standalone-visualize.sh | bash`.
• *Note: The standalone bootstrap requires `git` and `node`/`npm`.*

2. Install dependencies and build if needed (if running from source):

• `cd [detected-visualize-dir] && npm install && npm run build`

3. Run the visualizer:

• Production server: `cd [detected-visualize-dir] && npm run server [args]` (serves on port 3141)
• Development mode: `cd [detected-visualize-dir] && npm run dev [args]`
• Or if using the wrapper: `~/.pi/visualize.sh [args]`

4. If the user passed additional flags after `visualize`, forward them exactly as provided.

• Example: `/pi visualize --no-open --output /tmp/pi.html`

5. On success, report:

• server URL (default: http://localhost:3141)
• whether the browser was opened automatically
• session / warning counts from the server output

If the user says `/pi visualize` with no extra flags, use the default behavior:

• source: `~/.pi/decisions`
• server port: 3141
• browser: auto-open enabled

**After loading PI skill, always output the scene announcement (场景公示) so the user can see which mode and scene are active.**