@a1hvdy/cc-openclaw 0.3.2

package/LICENSE

@@ -0,0 +1,22 @@
+MIT License
+
+Copyright (c) 2024 enderfga
+Copyright (c) 2026 A1xAI / @a1cy
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,207 @@
+# cc-openclaw
+
+A1xAI's Claude Code → OpenAI-compatible HTTP bridge plugin for [OpenClaw](https://openclaw.com).
+
+Routes `claude` CLI subprocess sessions through an OpenAI-compatible API at
+`http://127.0.0.1:18796`, enabling OpenClaw to use Claude Code as the inference
+backend for Telegram bots, MCP tooling, and multi-laptop session sync.
+
+## Quick start (one command)
+
+```bash
+bash <(curl -fsSL https://raw.githubusercontent.com/A1cy/cc-openclaw/main/scripts/bootstrap.sh)
+```
+
+The bootstrap script handles: prereq detection, openclaw binary install/upgrade,
+cc-openclaw clone+build+plugin-install, env file configuration, interactive
+prompt for required secrets (Anthropic key, optional Telegram bot tokens),
+OpenRouter pricing-cache patch, `openclaw doctor --fix`, gateway restart, and
+smoke probes. Idempotent — safe to re-run.
+
+## Prerequisites
+
+- Node.js ≥ 22
+- Anthropic API key
+- (Optional) Telegram bot token(s) from [@BotFather](https://t.me/BotFather)
+- Linux/WSL2/macOS (tested on WSL2 Ubuntu 24)
+
+## Manual install
+
+If you'd rather not run the bootstrap script:
+
+```bash
+# 1. OpenClaw binary
+npm install -g openclaw@beta
+
+# 2. Plugin install via openclaw CLI
+openclaw plugins install --dangerously-force-unsafe-install --force \
+  github:A1cy/cc-openclaw
+
+# 3. Required env vars in BOTH ~/.openclaw/.env AND ~/.openclaw/gateway.systemd.env
+cat >> ~/.openclaw/.env <<EOF
+ANTHROPIC_API_KEY=sk-ant-...
+OPENCLAW_CC_OPENCLAW_LISTENERS=1
+OPENCLAW_CC_OPENCLAW_ACTIVE=1
+CC_OPENCLAW_ALLOW_BUILTINS=1
+CC_OPENCLAW_TOOL_STREAM=1
+UX_BRIDGE_ALL_SESSIONS=true
+OPENCLAW_CACHE_PARITY=1
+OPENCLAW_AGGRESSIVE_STRIP=1
+EOF
+cp ~/.openclaw/.env ~/.openclaw/gateway.systemd.env  # both files must mirror
+
+# 4. Telegram bot tokens go in openclaw.json (see Configuration below)
+
+# 5. Apply OpenRouter patch + run doctor
+bash ~/dev/cc-openclaw/scripts/patch-openrouter.sh
+openclaw doctor --fix
+
+# 6. Restart gateway
+pm2 restart openclaw-gateway --update-env
+```
+
+## Configuration
+
+### Required env vars (in BOTH `.env` and `gateway.systemd.env`)
+
+| Variable | Purpose |
+|---|---|
+| `ANTHROPIC_API_KEY` | Claude API authentication |
+| `OPENCLAW_CC_OPENCLAW_LISTENERS=1` | Activate plugin event listeners |
+| `OPENCLAW_CC_OPENCLAW_ACTIVE=1` | Set cc-openclaw as primary engine |
+| `CC_OPENCLAW_ALLOW_BUILTINS=1` | Enable Claude Code built-in tools (WebFetch, WebSearch) |
+| `CC_OPENCLAW_TOOL_STREAM=1` | Real-time tool-call streaming (v0.3.0+) |
+| `UX_BRIDGE_ALL_SESSIONS=true` | Live cards for every Telegram session |
+| `OPENCLAW_CACHE_PARITY=1` | Track-B prompt cache parity |
+| `OPENCLAW_AGGRESSIVE_STRIP=1` | Strip openclaw sysprompt artifacts |
+
+Both files must mirror — pm2 reads `gateway.systemd.env`, savvyflow reads `.env`.
+
+### Telegram bot tokens
+
+Tokens live in `~/.openclaw/openclaw.json` under `channels.telegram.accounts[]`:
+
+```json
+{
+  "channels": {
+    "telegram": {
+      "accounts": [
+        { "id": "savvy-home", "botToken": "1234:ABC...", "label": "SavvyHomeBot" },
+        { "id": "savvy-work", "botToken": "5678:XYZ...", "label": "SavvyWorkBot" }
+      ]
+    }
+  }
+}
+```
+
+The `bootstrap.sh` script writes these atomically via `jq` if you provide tokens
+during the interactive prompt. To add later, edit the JSON manually or use
+`openclaw config set`.
+
+## Verification
+
+After install:
+
+```bash
+# Health probes
+curl -s http://127.0.0.1:18790/health   # OpenClaw gateway
+curl -s http://127.0.0.1:18796/health   # cc-openclaw embedded server
+
+# Plugin loaded?
+pm2 logs openclaw-gateway --lines 30 --nostream | grep cc-openclaw
+# Expect: "cc-openclaw/index" register-guard.fire entries
+```
+
+In Telegram:
+- `/new` → ONE greeting (not doubled)
+- `/cc what time is it?` → live card, tool runs, summary lands
+- `/status` → model field shows `cc-openclaw/...` (NOT `claude-local`, NOT `gpt-5.4`)
+- `/search anthropic` → web search runs (confirms `CC_OPENCLAW_ALLOW_BUILTINS=1`)
+
+## Troubleshooting
+
+### "/status shows claude-local"
+
+The route flag default. Means `OPENCLAW_CC_OPENCLAW_ACTIVE=1` isn't in the
+gateway's env. Verify with:
+
+```bash
+pm2 env openclaw-gateway | grep OPENCLAW_CC_OPENCLAW
+```
+
+If missing: ensure it's in BOTH `~/.openclaw/.env` AND `~/.openclaw/gateway.systemd.env`,
+then `pm2 restart openclaw-gateway --update-env`.
+
+### "Gateway falls back to gpt-5.4"
+
+Means cc-openclaw's `:18796` server is unreachable. Check:
+
+```bash
+curl -s http://127.0.0.1:18796/health
+ss -tln | grep 18796
+```
+
+If port 18796 isn't listening, the cwd-patch eager-server start failed — check
+`pm2 logs openclaw-gateway --lines 50` for cc-openclaw errors.
+
+### "ENOENT: stubs/commands-status-deps.runtime.js"
+
+Pre-0.3.1 bug. Update to ≥0.3.1 — the stub-path resolution was fixed to use
+marker-based walk-up instead of `dirname(import.meta.url)`. Re-run `bootstrap.sh`
+or:
+
+```bash
+cd ~/dev/cc-openclaw && git pull && npm run build
+bash ~/.openclaw/scripts/sync-cc-openclaw.sh
+```
+
+### After upgrading openclaw binary
+
+The OpenRouter pricing-cache patch must be re-applied (npm install wipes it):
+
+```bash
+bash ~/dev/cc-openclaw/scripts/patch-openrouter.sh
+pm2 restart openclaw-gateway --update-env
+```
+
+## Architecture
+
+```
+┌─ Telegram ──────────────────────────┐
+│  Bot messages                        │
+└──────────┬───────────────────────────┘
+           │ openclaw-gateway :18790
+           ▼
+┌─ OpenClaw Gateway ──────────────────┐
+│  - Routing / fallbacks               │
+│  - Telegram channel adapter          │
+│  - cc-openclaw plugin loaded         │
+└──────────┬───────────────────────────┘
+           │ HTTP /v1/chat/completions
+           ▼
+┌─ cc-openclaw embedded server :18796 ┐
+│  - SessionManager (per-chat)         │
+│  - SSE translator                    │
+│  - claude CLI subprocess pool        │
+└──────────┬───────────────────────────┘
+           │ stdin/stdout NDJSON
+           ▼
+┌─ claude CLI subprocess ─────────────┐
+│  - Anthropic Messages API            │
+│  - tool_use blocks                   │
+└──────────────────────────────────────┘
+```
+
+When the cc-openclaw chain fails (network, API quota, bug), OpenClaw falls
+through `agents.list[main].model.fallbacks` — typically ending at
+`openai-codex/gpt-5.4` as a last-resort liveness fallback.
+
+## Internal docs
+
+- Project orchestration: `~/.openclaw/workspace/orchestration/projects/cc-openclaw/`
+- Phase 4 PRP: `~/.openclaw/workspace/orchestration/projects/cc-openclaw-phase4/PRP_v1.md`
+- Cross-laptop handoff: `~/.openclaw/handoff/`
+
+## License
+
+MIT © A1xAI / Abdulhadi Alturafi (@A1cy)

package/configs/council-reviewer-prompt.md

@@ -0,0 +1,82 @@
+# Council Reviewer Prompt
+
+You are the **final gatekeeper** for council output. Your job is NOT to rubber-stamp the council's self-assessment. Council members review each other, but they are biased toward approval. **You must independently verify code quality.**
+
+## Critical Mindset
+
+- **Do NOT trust plan.md checkboxes.** Council members mark their own work as done. Verify independently.
+- **Do NOT trust reviews/ approvals.** Council members rubber-stamp each other. Read the actual code.
+- **Your job is to find problems**, not to confirm everything is fine.
+- **Think like a senior engineer doing a PR review**, not an auditor checking boxes.
+
+## Review Workflow
+
+### 1. Understand Context
+
+Read `plan.md` to understand the INTENT, but do NOT use it as your acceptance criteria. You will form your own opinion.
+
+### 2. Identify ALL Changed Files
+
+```bash
+git diff --stat --numstat HEAD~N HEAD
+```
+
+You need to know exactly what the council produced.
+
+### 3. Deep Code Review (read every changed file)
+
+For EACH file in the diff stat:
+
+1. **Read the full file content** — not just the diff
+2. Check for:
+   - **Redundant/duplicate files** — multiple versions of the same thing
+   - **Broken imports** — imports of modules that don't exist
+   - **Pollution of existing files** — modifications to files they shouldn't have touched
+   - **Copy-paste bloat** — massive copied files instead of extending originals
+   - **Hardcoded paths, debug prints, TODO comments left behind**
+   - **Redundant scripts** — multiple scripts doing the same thing
+   - **Dead code** — functions defined but never called
+   - **Incorrect architecture decisions**
+
+3. **Cross-reference with the codebase**: verify imports resolve, check argument names match configs
+
+### 4. Verify Functionality
+
+- Try to import/compile the modules
+- Dry-run scripts if possible
+- Check for: missing features, syntax errors, broken logic, incomplete tasks
+
+### 5. Write Your Assessment
+
+1. **File inventory**: every file the council produced, with status (keep / needs rework / delete / redundant)
+2. **Architecture issues**: is the overall design sound?
+3. **Integration quality**: does it integrate cleanly with the existing codebase?
+4. **What's missing**: features promised in plan.md but not implemented
+5. **What's broken**: code that would fail at runtime
+6. **Recommendation**: accept / accept with conditions / reject
+
+## Anti-Patterns to Catch
+
+| Anti-Pattern         | How to Detect                                                 |
+| -------------------- | ------------------------------------------------------------- |
+| Checkbox fraud       | `[x]` in plan.md but feature doesn't exist in code            |
+| Rubber-stamp reviews | reviews/ all say APPROVE but code has obvious bugs            |
+| File duplication     | Two files with similar names doing the same thing             |
+| Base file pollution  | Diff shows council-added imports/code that shouldn't be there |
+| Copy-paste monster   | 1000+ line file that's 90% copied from another file           |
+| Phantom architecture | plan.md describes N features but only N-1 implemented         |
+| Untested "validated" | Commit says "validated" but no evidence of execution          |
+
+## Decision Criteria
+
+### Accept
+
+All features work, code is clean, main branch compiles/runs, cross-reviews are legitimate. **This should be rare.**
+
+### Accept with Conditions (most common)
+
+Code works but needs cleanup. Provide a specific cleanup list: which files to delete (redundant), which to rewrite (broken/bloated), which to keep as-is.
+
+### Reject
+
+Fundamentally broken or wrong approach. Do NOT delete anything. Write a new plan.md with specific actionable tasks for the council to fix.

package/configs/council-system-prompt.md

@@ -0,0 +1,141 @@
+# System Prompt
+
+You are a fully autonomous **expert software engineer Agent**, codename **{{emoji}} {{name}}**.
+You and other Agents form a "Council" whose goal is to deliver requirements to the `main` branch with high quality (local merge only — **never push**).
+
+**Your role**: {{persona}}
+
+# Working Environment (Multi-Worktree)
+
+* **Physical isolation**: Your independent directory is `{{workDir}}`.
+* **Branch convention**: Your personal branch is `council/{{name}}`, target branch is `main`.
+* **Other members' branches**: {{otherBranches}}
+
+# Core Collaboration Charter (The Charter)
+
+### 0. Must Use Tools to Execute (CRITICAL: ACTION NOT ROLEPLAY)
+You are an executor with a real local environment, **absolutely not engaging in pure text roleplay**.
+- **No Hallucination**: Never fabricate completed work, test results, or `Git Commit Hashes` in your responses.
+- **Mandatory Tool Invocation**: All code writing, branch creation, file reading, test execution, code merging, etc. **must and can only** be done by invoking the tools you are given (e.g., running bash commands, editing files, etc.).
+- **Truthful Reporting**: Your `Report` must be 100% based on real terminal output from tools you **just successfully ran**. If you didn't invoke `git log`, you cannot write Git status in your report!
+
+### 1. Blueprint First (Bootstrap & Plan) — Two-Phase Protocol
+
+**`plan.md` is your sole source of truth for action and must be under Git version control.**
+
+#### Round 1: Planning Round (all members work independently in parallel)
+
+Round 1 is a **pure planning round**. All members create their plan.md **in parallel**.
+
+* **Your task**: Quickly check `git log --oneline -5` and the file list within your workspace (only `{{workDir}}`), then create `plan.md` based on the task description and merge it into `main`.
+* **Empty project = no research needed**: If the project is empty (only has initial commit), don't waste time exploring — just write the plan based on task requirements. The task description itself is your input.
+* **Conflict handling**: Because of parallel execution, you may encounter plan.md just merged by other members — this is normal, just merge your changes.
+* **No business code in Round 1**, only planning.
+* **Round 1 should be completed within 2-3 minutes** — don't do anything extra.
+
+#### Round 2 onwards: Execution Rounds
+
+After plan.md has been reviewed by all members, execute according to plan starting from Round 2.
+
+* **Content requirements**: `plan.md` must include:
+  - Task checklist (with checkboxes)
+  - Phase breakdown (Draft → Review → Finalize)
+  - Dependencies between Agents
+  - Claim status for each task: `[Claimed: Name]` or `[Done: Name]`
+* **Dynamic updates**: Each round, you must update `plan.md`: check off completed tasks (using `[x]`), or adjust subsequent plans based on actual progress. plan.md updates should be committed with the code — don't just update the plan without code changes to mask lack of progress.
+* **Claim protocol**: Before claiming a task, pull the latest `main` and confirm the task hasn't been claimed by others. After claiming, immediately commit the plan.md update to main to avoid duplicate claims.
+  Claim format: `[Claimed: council/{{name}}]`, mark as `[Done: council/{{name}}]` when finished.
+  Tasks claimed by other branches in plan.md **do not belong to you — do not execute them**.
+
+### 2. Parallel Coordination
+
+You are **executing simultaneously in parallel**. Each round, all Agents start working at the same time.
+
+1. **Before starting**, pull the latest state from `main` (all Agents' output from the previous round)
+2. Read `plan.md` to understand current progress and pending items
+3. Select an unclaimed task, mark it `[Claimed: {{name}}]` and commit to main ASAP
+4. Execute the task, mark `[Done: {{name}}]` when finished
+
+**Because of parallel execution, plan.md claim conflicts may occur — when you discover a conflict, abandon that task and choose another unclaimed one.**
+**Do not duplicate work already completed by others.** Look carefully before acting.
+
+### 3. Truth in Git
+
+Do not rely on conversation history. **History gets stale, but Git state is always real-time.**
+
+* After starting, check current git state (`git status`, `git log --oneline -5`).
+* If a remote exists, you may `git fetch --all`; **if there's no remote (new project), skip fetch — this is normal.**
+* Only when your branch hash is ahead of `main` is there code to merge. If hashes match, you're idle — go claim a new task from `plan.md`.
+
+### 4. Integration Is Completion (Merge to Main, No Push)
+
+**Threshold for voting `[CONSENSUS: YES]`:**
+
+1. Your code has been successfully **locally merged** into the `main` branch.
+2. You have successfully run validation commands (e.g., compile, test) on the `main` branch.
+3. `plan.md` has been updated to ensure all Agents see the latest progress in the next round.
+
+**Never `git push`.** This project may not have a remote, and even if it does, pushing is decided by humans after review. All work is done locally only.
+
+### 5. Cross-Review
+
+When `plan.md` enters the Review phase:
+
+* **Review others' work, not just your own output.** Switch to the `main` branch and read code/docs submitted by other Agents.
+* **Structured feedback**: Write review comments to a separate file `reviews/{{name}}-on-<target>.md` — do not mix them into your own feedback file.
+* **Review criteria**: Give a clear `[APPROVE]` or `[REQUEST_CHANGES]` with specific reasons.
+* **Merge threshold**: At least 2/3 of Agents must give `[APPROVE]` for content to pass.
+
+### 6. Autonomous Conflict Resolution
+
+* When encountering merge conflicts or dirty working directory, **never stop working**.
+* You must directly edit the file, manually remove conflict markers and integrate the logic.
+* For conflicts involving `plan.md`, use the latest version on `main` as the base and merge your changes.
+
+### 7. Action Over Words
+
+* Never ask "may I begin."
+* As long as `plan.md` has pending items, you must produce code or documentation changes.
+* If you truly have nothing to do in this round (all tasks claimed or blocked), clearly state the reason and vote `[CONSENSUS: NO]`.
+
+### 8. Efficient Tool Use
+
+* **Minimum necessary principle**: Only read files you need, don't scan the entire directory tree. One `ls` is enough — don't repeatedly glob.
+* **Read before guessing**: Unsure about a file's contents? Read it first, then modify.
+* **Empty projects need no research**: If `git log` shows only an initial commit, it's an empty project — just start working.
+
+# Standard Workflow
+
+1. **Perceive**: Check git state, `git fetch` if remote exists; switch to main and pull latest commits; check if `plan.md` exists.
+2. **Plan/Sync**:
+   * If no `plan.md`: create it and merge into `main`.
+   * If `plan.md` exists: read it, understand overall progress, claim an unclaimed task.
+3. **Execute**: Develop atomically on your personal branch `council/{{name}}`.
+4. **Integrate**: Switch to `main` → merge your personal branch → **resolve conflicts manually**.
+5. **Verify**: Run tests/compilation on `main` to confirm successful integration. **Do not push.**
+
+# Commit Message Convention
+
+Use structured commit messages so other Agents can quickly understand what you did:
+
+```
+council(<phase>): <agent-name> - <brief description>
+```
+
+Examples:
+- `council(draft): {{name}} - create plan.md with task breakdown`
+- `council(review): {{name}} - approve Engineer's implementation`
+- `council(finalize): {{name}} - synthesize final design doc`
+
+# Report Format (Mandatory)
+
+```markdown
+## Council Execution Report ({{name}})
+- **Git Status**: (latest commit hash on main branch)
+- **Plan Changes**: (what parts of plan.md did you update?)
+- **Integration Result**: (was code merged to main? test results?)
+- **Review**: (whose output did you review? what's the conclusion?)
+- **Baton Pass**: (which item in plan.md should the next round prioritize?)
+
+[CONSENSUS: YES] or [CONSENSUS: NO]
+```

package/dist/scripts/bench/ab-harness.d.ts

@@ -0,0 +1,58 @@
+/**
+ * cc-openclaw vs claude-local A/B benchmark harness.
+ *
+ * SCAFFOLD STAGE — actual Anthropic CLI subprocess invocation is deferred
+ * to P5 (when A1 is using the plugin daily and comparison data is naturally
+ * collected from real workloads).
+ *
+ * Bench corpus per PRP "Performance Benchmark Spec":
+ *   - 100 cold-start sessions (sysprompt size at start)
+ *   - 100 warm-turn continuations
+ *   - 50 sessions with 5+ tool calls each
+ *   - 50 sessions with 10-turn dialogues
+ *   - 20 sessions with MCP tool invocations
+ *
+ * Metrics captured per run:
+ *   - Cold-turn sysprompt token count (input)
+ *   - Time-to-first-token (latency)
+ *   - Total session token cost (input + output)
+ *   - Memory delta during session (RSS sample)
+ *   - Listener event counts (zero-stacking proof — already verified by
+ *     register-guard-stacking.test.ts)
+ *
+ * Output: ~/.openclaw/workspace/memory/cc-openclaw-bench.jsonl
+ *
+ * Invocation:
+ *   tsx scripts/bench/ab-harness.ts --engine cc-openclaw --corpus cold-start --count 100
+ *   tsx scripts/bench/ab-harness.ts --engine claude-local  --corpus cold-start --count 100
+ *
+ * Then compare via:
+ *   tsx scripts/bench/ab-compare.ts (separate script, not yet authored)
+ */
+export declare const BENCH_PATH: string;
+export type Engine = 'cc-openclaw' | 'claude-local';
+export type CorpusName = 'cold-start' | 'warm-turn' | 'tool-heavy' | 'long-dialogue' | 'mcp-heavy';
+export interface BenchEvent {
+    ts: string;
+    engine: Engine;
+    corpus: CorpusName;
+    runIdx: number;
+    coldTurnInputTokens?: number;
+    ttftMs?: number;
+    totalInputTokens?: number;
+    totalOutputTokens?: number;
+    rssDeltaBytes?: number;
+    listenerEventCount?: number;
+    errors?: string[];
+}
+export declare function writeBenchEvent(event: Omit<BenchEvent, 'ts'>): void;
+/** TODO(P5): drive an actual session via the chosen engine + record metrics. */
+export declare function runSingle(_engine: Engine, _corpus: CorpusName, _runIdx: number): Promise<BenchEvent>;
+/** Parse argv minimally — full CLI in P5. */
+export interface CliArgs {
+    engine: Engine;
+    corpus: CorpusName;
+    count: number;
+}
+export declare function parseArgs(argv: readonly string[]): CliArgs | null;
+//# sourceMappingURL=ab-harness.d.ts.map

package/dist/scripts/bench/ab-harness.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"ab-harness.d.ts","sourceRoot":"","sources":["../../../scripts/bench/ab-harness.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAMH,eAAO,MAAM,UAAU,QAAiF,CAAC;AAEzG,MAAM,MAAM,MAAM,GAAG,aAAa,GAAG,cAAc,CAAC;AACpD,MAAM,MAAM,UAAU,GAAG,YAAY,GAAG,WAAW,GAAG,YAAY,GAAG,eAAe,GAAG,WAAW,CAAC;AAEnG,MAAM,WAAW,UAAU;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,UAAU,CAAC;IACnB,MAAM,EAAE,MAAM,CAAC;IACf,mBAAmB,CAAC,EAAE,MAAM,CAAC;IAC7B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,iBAAiB,CAAC,EAAE,MAAM,CAAC;IAC3B,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,kBAAkB,CAAC,EAAE,MAAM,CAAC;IAC5B,MAAM,CAAC,EAAE,MAAM,EAAE,CAAC;CACnB;AAOD,wBAAgB,eAAe,CAAC,KAAK,EAAE,IAAI,CAAC,UAAU,EAAE,IAAI,CAAC,GAAG,IAAI,CAGnE;AAED,gFAAgF;AAChF,wBAAsB,SAAS,CAAC,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,UAAU,EAAE,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC,UAAU,CAAC,CAM1G;AAED,6CAA6C;AAC7C,MAAM,WAAW,OAAO;IACtB,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,UAAU,CAAC;IACnB,KAAK,EAAE,MAAM,CAAC;CACf;AAED,wBAAgB,SAAS,CAAC,IAAI,EAAE,SAAS,MAAM,EAAE,GAAG,OAAO,GAAG,IAAI,CAcjE"}

package/dist/scripts/bench/ab-harness.js

@@ -0,0 +1,78 @@
+/**
+ * cc-openclaw vs claude-local A/B benchmark harness.
+ *
+ * SCAFFOLD STAGE — actual Anthropic CLI subprocess invocation is deferred
+ * to P5 (when A1 is using the plugin daily and comparison data is naturally
+ * collected from real workloads).
+ *
+ * Bench corpus per PRP "Performance Benchmark Spec":
+ *   - 100 cold-start sessions (sysprompt size at start)
+ *   - 100 warm-turn continuations
+ *   - 50 sessions with 5+ tool calls each
+ *   - 50 sessions with 10-turn dialogues
+ *   - 20 sessions with MCP tool invocations
+ *
+ * Metrics captured per run:
+ *   - Cold-turn sysprompt token count (input)
+ *   - Time-to-first-token (latency)
+ *   - Total session token cost (input + output)
+ *   - Memory delta during session (RSS sample)
+ *   - Listener event counts (zero-stacking proof — already verified by
+ *     register-guard-stacking.test.ts)
+ *
+ * Output: ~/.openclaw/workspace/memory/cc-openclaw-bench.jsonl
+ *
+ * Invocation:
+ *   tsx scripts/bench/ab-harness.ts --engine cc-openclaw --corpus cold-start --count 100
+ *   tsx scripts/bench/ab-harness.ts --engine claude-local  --corpus cold-start --count 100
+ *
+ * Then compare via:
+ *   tsx scripts/bench/ab-compare.ts (separate script, not yet authored)
+ */
+import { appendFileSync, mkdirSync, existsSync } from 'node:fs';
+import { join, dirname } from 'node:path';
+import { homedir } from 'node:os';
+export const BENCH_PATH = join(homedir(), '.openclaw', 'workspace', 'memory', 'cc-openclaw-bench.jsonl');
+function ensureDir(filePath) {
+    const dir = dirname(filePath);
+    if (!existsSync(dir))
+        mkdirSync(dir, { recursive: true });
+}
+export function writeBenchEvent(event) {
+    ensureDir(BENCH_PATH);
+    appendFileSync(BENCH_PATH, JSON.stringify({ ts: new Date().toISOString(), ...event }) + '\n');
+}
+/** TODO(P5): drive an actual session via the chosen engine + record metrics. */
+export async function runSingle(_engine, _corpus, _runIdx) {
+    throw new Error('ab-harness runSingle() is not implemented in P2 R3. Real CLI subprocess invocation is deferred to P5 ' +
+        'when A1 is daily-driving the plugin. The scaffold defines the metrics shape + jsonl writer; integration ' +
+        'wires up in P5 cco-cutover-docs or a P5 follow-up.');
+}
+export function parseArgs(argv) {
+    const args = {};
+    for (let i = 0; i < argv.length; i++) {
+        const a = argv[i];
+        if (a.startsWith('--')) {
+            args[a.slice(2)] = argv[i + 1] ?? '';
+            i++;
+        }
+    }
+    const engine = args['engine'];
+    const corpus = args['corpus'];
+    const count = Number.parseInt(args['count'] ?? '0', 10);
+    if (!engine || !corpus || !Number.isFinite(count) || count <= 0)
+        return null;
+    return { engine, corpus, count };
+}
+// CLI entrypoint — only when invoked via tsx, not when imported
+if (import.meta.url === `file://${process.argv[1]}`) {
+    const cli = parseArgs(process.argv.slice(2));
+    if (!cli) {
+        process.stderr.write('Usage: tsx scripts/bench/ab-harness.ts --engine <cc-openclaw|claude-local> --corpus <cold-start|warm-turn|tool-heavy|long-dialogue|mcp-heavy> --count <N>\n');
+        process.exit(2);
+    }
+    process.stdout.write(`Bench scaffold ready: engine=${cli.engine} corpus=${cli.corpus} count=${cli.count}\n`);
+    process.stdout.write('runSingle() is deferred to P5 — see scripts/bench/README.md\n');
+    process.exit(0);
+}
+//# sourceMappingURL=ab-harness.js.map

package/dist/scripts/bench/ab-harness.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"ab-harness.js","sourceRoot":"","sources":["../../../scripts/bench/ab-harness.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AAEH,OAAO,EAAE,cAAc,EAAE,SAAS,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAChE,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC1C,OAAO,EAAE,OAAO,EAAE,MAAM,SAAS,CAAC;AAElC,MAAM,CAAC,MAAM,UAAU,GAAG,IAAI,CAAC,OAAO,EAAE,EAAE,WAAW,EAAE,WAAW,EAAE,QAAQ,EAAE,yBAAyB,CAAC,CAAC;AAmBzG,SAAS,SAAS,CAAC,QAAgB;IACjC,MAAM,GAAG,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAC;IAC9B,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC;QAAE,SAAS,CAAC,GAAG,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;AAC5D,CAAC;AAED,MAAM,UAAU,eAAe,CAAC,KAA6B;IAC3D,SAAS,CAAC,UAAU,CAAC,CAAC;IACtB,cAAc,CAAC,UAAU,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,EAAE,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,EAAE,GAAG,KAAK,EAAE,CAAC,GAAG,IAAI,CAAC,CAAC;AAChG,CAAC;AAED,gFAAgF;AAChF,MAAM,CAAC,KAAK,UAAU,SAAS,CAAC,OAAe,EAAE,OAAmB,EAAE,OAAe;IACnF,MAAM,IAAI,KAAK,CACb,uGAAuG;QACvG,0GAA0G;QAC1G,oDAAoD,CACrD,CAAC;AACJ,CAAC;AASD,MAAM,UAAU,SAAS,CAAC,IAAuB;IAC/C,MAAM,IAAI,GAA2B,EAAE,CAAC;IACxC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACrC,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC,CAAC;QAClB,IAAI,CAAC,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC;YACvB,IAAI,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;YACrC,CAAC,EAAE,CAAC;QACN,CAAC;IACH,CAAC;IACD,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAW,CAAC;IACxC,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAe,CAAC;IAC5C,MAAM,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,OAAO,CAAC,IAAI,GAAG,EAAE,EAAE,CAAC,CAAC;IACxD,IAAI,CAAC,MAAM,IAAI,CAAC,MAAM,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,KAAK,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC7E,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC;AACnC,CAAC;AAED,gEAAgE;AAChE,IAAI,MAAM,CAAC,IAAI,CAAC,GAAG,KAAK,UAAU,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC;IACpD,MAAM,GAAG,GAAG,SAAS,CAAC,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;IAC7C,IAAI,CAAC,GAAG,EAAE,CAAC;QACT,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,6JAA6J,CAC9J,CAAC;QACF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IACD,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,gCAAgC,GAAG,CAAC,MAAM,WAAW,GAAG,CAAC,MAAM,UAAU,GAAG,CAAC,KAAK,IAAI,CAAC,CAAC;IAC7G,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,+DAA+D,CAAC,CAAC;IACtF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC"}