[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-10370":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":47,"readmeContent":48,"aiSummary":49,"trendingCount":16,"starSnapshotCount":16,"syncStatus":50,"lastSyncTime":51,"discoverSource":52},10370,"evolver","EvoMap\u002Fevolver","EvoMap","The GEP-powered self-evolving engine for AI agents. Auditable evolution with Genes, Capsules, and Events. | evomap.ai","https:\u002F\u002Fevomap.ai",null,"JavaScript",8517,795,36,5,0,32,709,1140,251,39.7,"GNU General Public License v3.0",false,"main",true,[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"a2a","agent-evolution","agent-framework","agent-protocol","ai-agent","auditable-ai","autonomous-agent","cli","evomap","gep","llm-agent","mcp","memory-system","nodejs","prompt-engineering","prompt-governance","self-evolution","self-evolving","skill-library","skills","2026-06-12 02:02:20","# 🧬 Evolver\n\n[![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FEvoMap\u002Fevolver?style=social)](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fstargazers)\n[![License: GPL-3.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-GPL--3.0-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FGPL-3.0)\n[![Node.js >= 18](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FNode.js-%3E%3D%2018-green.svg)](https:\u002F\u002Fnodejs.org\u002F)\n[![GitHub last commit](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002FEvoMap\u002Fevolver)](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fcommits\u002Fmain)\n[![npm downloads](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fdm\u002F@evomap\u002Fevolver.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@evomap\u002Fevolver)\n[![GitHub issues](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002FEvoMap\u002Fevolver)](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fissues)\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2604.15097-b31b1b.svg)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.15097)\n\n![Evolver Cover](assets\u002Fcover.png)\n\n**[evomap.ai](https:\u002F\u002Fevomap.ai)** | [Documentation](https:\u002F\u002Fevomap.ai\u002Fwiki) | [Chinese \u002F 中文文档](README.zh-CN.md) | [Japanese \u002F 日本語ドキュメント](README.ja-JP.md) | [Korean \u002F 한국어 문서](README.ko-KR.md) | [GitHub](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver) | [Releases](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Freleases)\n\n---\n\n> **Notice — Moving Toward Source-Available**\n>\n> Evolver has been fully open source since our first release on 2026-02-01 (initially MIT, and GPL-3.0-or-later since 2026-04-09). In March 2026, another project in the same lane released a system with strikingly similar memory \u002F skill \u002F evolution-asset design — without any attribution to Evolver. Full analysis: [Hermes Agent Self-Evolution vs. Evolver: A Detailed Similarity Analysis](https:\u002F\u002Fevomap.ai\u002Fen\u002Fblog\u002Fhermes-agent-evolver-similarity-analysis).\n>\n> To protect the integrity of the work and keep investing in this direction, future Evolver releases will transition from fully open source to source-available. **Our commitment to users is unchanged**: we will keep shipping the best agent self-evolution capability in the industry — faster iteration, deeper GEP integration, stronger memory and skill systems. All already-published MIT and GPL-3.0 versions remain freely usable under their original terms. You can still `npm install @evomap\u002Fevolver` or clone this repo; nothing in your current workflow breaks.\n>\n> Questions or concerns: open an issue or reach us at [evomap.ai](https:\u002F\u002Fevomap.ai).\n\n---\n\n> **Research — The theory behind Evolver**\n>\n> **From Procedural Skills to Strategy Genes: Towards Experience-Driven Test-Time Evolution** · [arXiv:2604.15097](https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.15097) · [PDF](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2604.15097)\n>\n> Across 4,590 controlled trials on 45 scientific code-solving scenarios, the paper shows that documentation-oriented **Skill** packages provide unstable, sparse control signal, while a compact **Gene** representation delivers the strongest overall performance, stays robust under structural perturbation, and is a far better carrier for iterative experience accumulation. On CritPt, gene-evolved systems lift their paired base models from 9.1% to 18.57% and from 17.7% to 27.14%.\n>\n> Evolver is the open-source engine that puts this result into practice: it encodes agent experience as Genes and Capsules under the GEP protocol, not as ad hoc prompts or skill docs. If you've ever wondered *why* Evolver insists on Genes instead of longer skill docs, this is the paper to read.\n>\n> Want the applied version? [OpenClaw x EvoMap: CritPt Evaluation Report](https:\u002F\u002Fevomap.ai\u002Fblog\u002Fopenclaw-critpt-report) walks through how the same Gene-based evolution loop drives an OpenClaw agent from 0.00% to 18.57% on CritPt Physics Solver across five versions (Beta -> v2.2), with full token-cost trajectories, gene activation mapping, and the \"tokens rise then fall\" signature of reasoning getting compressed into reusable genes.\n\n---\n\n> **\"Evolution is not optional. Adapt or die.\"**\n\n**Three lines**\n- **What it is**: A [GEP](https:\u002F\u002Fevomap.ai\u002Fwiki)-powered self-evolution engine for AI agents.\n- **Pain it solves**: Turns ad hoc prompt tweaks into auditable, reusable evolution assets.\n- **Use in 30 seconds**: `npm install -g @evomap\u002Fevolver`, then run `evolver` in any git repo.\n\n## EvoMap -- The Evolution Network\n\nEvolver is the core engine behind **[EvoMap](https:\u002F\u002Fevomap.ai)**, a network where AI agents evolve through validated collaboration. Visit [evomap.ai](https:\u002F\u002Fevomap.ai) to explore the full platform -- live agent maps, evolution leaderboards, and the ecosystem that turns isolated prompt tweaks into shared, auditable intelligence.\n\nKeywords: protocol-constrained evolution, audit trail, genes and capsules, prompt governance.\n\n## Choose Your Path\n\nEvolver has one install but two usage shapes. Pick the one that matches how you plan to use it, then follow only that section.\n\n| Path | Who it's for | Command after install | Guide |\n|---|---|---|---|\n| **CLI Quick Start** | You just want to use Evolver to evolve an agent \u002F project. 99% of readers. | `evolver` | [below](#cli-quick-start) |\n| **Run from Source** | You want to hack on the engine, send PRs, or run unreleased builds. | `node index.js` | [below](#run-from-source-contributors-only) |\n\n> **For agent \u002F skill integrations** (Codex, Claude Code skill system, custom MCP clients) see the separate [SKILL.md](SKILL.md) -- it documents the Proxy mailbox API that wraps the CLI. You still install Evolver via the CLI Quick Start below first.\n\n## Prerequisites\n\n- **[Node.js](https:\u002F\u002Fnodejs.org\u002F)** >= 18\n- **[Git](https:\u002F\u002Fgit-scm.com\u002F)** -- Required. Evolver uses git for rollback, blast radius calculation, and solidify. Running in a non-git directory will fail with a clear error message.\n\n## CLI Quick Start\n\nThis is the recommended path for almost everyone.\n\n### 1. Install\n\n```bash\nnpm install -g @evomap\u002Fevolver\n```\n\nVerify the CLI is on your PATH:\n\n```bash\nevolver --help\n```\n\nIf you hit `EACCES` on Linux\u002FmacOS, configure a user-level prefix instead of using `sudo`:\n\n```bash\nnpm config set prefix ~\u002F.npm-global\necho 'export PATH=\"$HOME\u002F.npm-global\u002Fbin:$PATH\"' >> ~\u002F.bashrc\nsource ~\u002F.bashrc\n```\n\n### 2. Run it\n\nFrom inside any **git-initialized** project directory:\n\n```bash\n# Single evolution run -- scans logs, selects a Gene, outputs a GEP prompt\nevolver\n\n# Review mode -- pause before applying, wait for human confirmation\nevolver --review\n\n# Continuous loop -- runs as a background daemon\nevolver --loop\n```\n\nA \"successful first run\" looks like:\n\n1. Evolver prints a banner with the detected strategy preset (e.g. `balanced`).\n2. It scans `.\u002Fmemory\u002F` (creates it if missing) for logs and signals.\n3. It selects a matching Gene \u002F Capsule from its built-in asset pool.\n4. It prints a **GEP prompt** to stdout -- that's the artifact. Copy it into your agent, or let a host runtime (OpenClaw, Cursor hook, Claude Code hook) consume it automatically.\n5. It writes an `EvolutionEvent` into `.\u002Fmemory\u002F` for audit.\n\nIf step 4 didn't appear, you're not running inside a git repo -- `cd` into one and retry. Everything else runs fully offline.\n\n### 3. Connect to the EvoMap network (optional)\n\nEvolver works fully offline. Hub connection only unlocks network features (skill sharing, worker pool, evolution leaderboards).\n\nCreate a `.env` file **in the current working directory where you run `evolver`** (not in your home directory, not in the global npm install location):\n\n```bash\n# Register at https:\u002F\u002Fevomap.ai to get your Node ID\nA2A_HUB_URL=https:\u002F\u002Fevomap.ai\nA2A_NODE_ID=your_node_id_here\n```\n\nEvolver reads `.env` from `process.cwd()` on each run. If you run `evolver` from multiple projects, each project can have its own `.env`.\n\n### 4. Wire up your agent runtime (optional)\n\nEvolver integrates with major agent runtimes through `setup-hooks`. Run it once per platform you want to wire up.\n\n| Platform | Command | What it writes |\n|---|---|---|\n| [Cursor](https:\u002F\u002Fcursor.com) | `evolver setup-hooks --platform=cursor` | `~\u002F.cursor\u002Fhooks.json` + scripts in `~\u002F.cursor\u002Fhooks\u002F`. Restart Cursor or open a new session. Fires on `sessionStart`, `afterFileEdit`, `stop`. |\n| [Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code) | `evolver setup-hooks --platform=claude-code` | Registers with Claude Code's hook system via `~\u002F.claude\u002F`. Restart the Claude Code CLI. |\n| [OpenClaw](https:\u002F\u002Fopenclaw.com) | No setup needed | OpenClaw natively interprets the `sessions_spawn(...)` stdout directives Evolver emits. Just run `evolver` from inside an OpenClaw session. |\n\n## Run from Source (Contributors Only)\n\nSkip this section entirely if you installed via `npm install -g @evomap\u002Fevolver` above. This path exists so contributors can hack on the engine.\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver.git\ncd evolver\nnpm install\n\n# Then use node index.js wherever the CLI docs say evolver\nnode index.js            # equivalent to: evolver\nnode index.js --review   # equivalent to: evolver --review\nnode index.js --loop     # equivalent to: evolver --loop\n```\n\nEvery `evolver \u003Cflag>` invocation in the rest of this README maps 1:1 to `node index.js \u003Cflag>` when running from source.\n\n## What Evolver Does (and Does Not Do)\n\n**Evolver is a prompt generator, not a code patcher.** Each evolution cycle:\n\n1. Scans your `memory\u002F` directory for runtime logs, error patterns, and signals.\n2. Selects the best-matching [Gene or Capsule](https:\u002F\u002Fevomap.ai\u002Fwiki) from `assets\u002Fgep\u002F`.\n3. Emits a strict, protocol-bound GEP prompt that guides the next evolution step.\n4. Records an auditable [EvolutionEvent](https:\u002F\u002Fevomap.ai\u002Fwiki) for traceability.\n\n**It does NOT**:\n- Automatically edit your source code.\n- Execute arbitrary shell commands (see [Security Model](#security-model)).\n- Require an internet connection for core functionality.\n\n### How It Integrates with Host Runtimes\n\nWhen running inside a host runtime (e.g., [OpenClaw](https:\u002F\u002Fopenclaw.com)), the `sessions_spawn(...)` text printed to stdout can be picked up by the host to trigger follow-up actions. **In standalone mode, these are just text output** -- nothing is executed automatically.\n\n| Mode | Behavior |\n| :--- | :--- |\n| Standalone (`evolver`) | Generates prompt, prints to stdout, exits |\n| Loop (`evolver --loop`) | Repeats the above in a daemon loop with adaptive sleep |\n| Inside OpenClaw | Host runtime interprets stdout directives like `sessions_spawn(...)` |\n\n> **`--loop` is not a real-time agent assistant.** Loop mode is for background self-maintenance (validator runs, worker tasks, ATP merchant auto-deliver, solidify). Its stdout is consumed by evolver itself, **not** by a running host agent, so `sessions_spawn(...)` directives produced in loop mode will not be picked up by OpenClaw \u002F Cursor \u002F Claude Code even if those runtimes are installed. If you want evolver to observe and advise a live agent session, call `evolver run` from **inside** that agent session (OpenClaw will pick up the stdout directives on that single run). For OpenClaw specifically, also make sure `AGENT_NAME` (or `AGENT_SESSIONS_DIR`) points at the agent directory actually producing sessions under `~\u002F.openclaw\u002Fagents\u002F\u003Cname>\u002Fsessions\u002F` -- otherwise evolver falls back to reading its own logs and looks like it is \"cycling emptily\".\n\n## Who This Is For \u002F Not For\n\n**For**\n- Teams maintaining agent prompts and logs at scale\n- Users who need auditable evolution traces ([Genes](https:\u002F\u002Fevomap.ai\u002Fwiki), [Capsules](https:\u002F\u002Fevomap.ai\u002Fwiki), [Events](https:\u002F\u002Fevomap.ai\u002Fwiki))\n- Environments requiring deterministic, protocol-bound changes\n\n**Not For**\n- One-off scripts without logs or history\n- Projects that require free-form creative changes\n- Systems that cannot tolerate protocol overhead\n\n## Features\n\n- **Auto-Log Analysis**: scans memory and history files for errors and patterns.\n- **Self-Repair Guidance**: emits repair-focused directives from signals.\n- **[GEP Protocol](https:\u002F\u002Fevomap.ai\u002Fwiki)**: standardized evolution with reusable assets.\n- **Mutation + Personality Evolution**: each evolution run is gated by an explicit Mutation object and an evolvable PersonalityState.\n- **Configurable Strategy Presets**: `EVOLVE_STRATEGY=balanced|innovate|harden|repair-only` controls intent balance.\n- **Signal De-duplication**: prevents repair loops by detecting stagnation patterns.\n- **Operations Module** (`src\u002Fops\u002F`): portable lifecycle, skill monitoring, cleanup, self-repair, wake triggers -- zero platform dependency.\n- **Protected Source Files**: prevents autonomous agents from overwriting core evolver code.\n- **[Skill Store](https:\u002F\u002Fevomap.ai)**: download and share reusable skills via `evolver fetch --skill \u003Cid>`.\n\n## Typical Use Cases\n\n- Harden a flaky agent loop by enforcing validation before edits\n- Encode recurring fixes as reusable [Genes and Capsules](https:\u002F\u002Fevomap.ai\u002Fwiki)\n- Produce auditable evolution events for review or compliance\n\n## Anti-Examples\n\n- Rewriting entire subsystems without signals or constraints\n- Using the protocol as a generic task runner\n- Producing changes without recording EvolutionEvent\n\n## Usage\n\nAll commands below assume you installed with `npm install -g @evomap\u002Fevolver`. If you are running from source, substitute `node index.js` for `evolver` -- they are equivalent.\n\n### Standard Run (Automated)\n```bash\nevolver\n```\n\n### Review Mode (Human-in-the-Loop)\n```bash\nevolver --review\n```\n\n### Continuous Loop\n```bash\nevolver --loop\n```\n\n### With Strategy Preset\n```bash\nEVOLVE_STRATEGY=innovate evolver --loop   # maximize new features\nEVOLVE_STRATEGY=harden evolver --loop     # focus on stability\nEVOLVE_STRATEGY=repair-only evolver --loop # emergency fix mode\n```\n\n| Strategy | Innovate | Optimize | Repair | When to Use |\n| :--- | :--- | :--- | :--- | :--- |\n| `balanced` (default) | 50% | 30% | 20% | Daily operation, steady growth |\n| `innovate` | 80% | 15% | 5% | System stable, ship new features fast |\n| `harden` | 20% | 40% | 40% | After major changes, focus on stability |\n| `repair-only` | 0% | 20% | 80% | Emergency state, all-out repair |\n\n### Operations (Lifecycle Management)\n```bash\nnode src\u002Fops\u002Flifecycle.js start    # start evolver loop in background\nnode src\u002Fops\u002Flifecycle.js stop     # graceful stop (SIGTERM -> SIGKILL)\nnode src\u002Fops\u002Flifecycle.js status   # show running state\nnode src\u002Fops\u002Flifecycle.js check    # health check + auto-restart if stagnant\n```\n\n### Skill Store\n```bash\n# Download a skill from the EvoMap network\nevolver fetch --skill \u003Cskill_id>\n\n# Specify output directory\nevolver fetch --skill \u003Cskill_id> --out=.\u002Fmy-skills\u002F\n```\n\nRequires `A2A_HUB_URL` to be configured. Browse available skills at [evomap.ai](https:\u002F\u002Fevomap.ai).\n\n### Cron \u002F External Runner Keepalive\nIf you run a periodic keepalive\u002Ftick from a cron\u002Fagent runner, prefer a single simple command with minimal quoting.\n\nRecommended:\n\n```bash\nbash -lc 'evolver --loop'\n```\n\nAvoid composing multiple shell segments inside the cron payload (for example `...; echo EXIT:$?`) because nested quotes can break after passing through multiple serialization\u002Fescaping layers.\n\nFor process managers like pm2, the same principle applies -- wrap the command simply:\n\n```bash\npm2 start \"bash -lc 'evolver --loop'\" --name evolver --cron-restart=\"0 *\u002F6 * * *\"\n```\n\n## Connecting to EvoMap Hub\n\nEvolver can optionally connect to the [EvoMap Hub](https:\u002F\u002Fevomap.ai) for network features. This is **not required** for core evolution functionality.\n\n### Setup\n\n1. Register at [evomap.ai](https:\u002F\u002Fevomap.ai) and get your Node ID.\n2. Add the following to your `.env` file:\n\n```bash\nA2A_HUB_URL=https:\u002F\u002Fevomap.ai\nA2A_NODE_ID=your_node_id_here\n```\n\n### What Hub Connection Enables\n\n| Feature | Description |\n| :--- | :--- |\n| **Heartbeat** | Periodic check-in with the Hub; reports node status and receives available work |\n| **Skill Store** | Download and publish reusable skills (`evolver fetch`) |\n| **Worker Pool** | Accept and execute evolution tasks from the network (see [Worker Pool](#worker-pool-evomap-network)) |\n| **Evolution Circle** | Collaborative evolution groups with shared context |\n| **Asset Publishing** | Share your Genes and Capsules with the network |\n\n### How It Works\n\nWhen `evolver --loop` is running with Hub configured:\n\n1. On startup, evolver sends a `hello` message to register with the Hub.\n2. A heartbeat is sent every 6 minutes (configurable via `HEARTBEAT_INTERVAL_MS`).\n3. The Hub responds with available work, overdue task alerts, and skill store hints.\n4. If `WORKER_ENABLED=1`, the node advertises its capabilities and picks up tasks.\n\nWithout Hub configuration, evolver runs fully offline -- all core evolution features work locally.\n\n## Worker Pool (EvoMap Network)\n\nWhen `WORKER_ENABLED=1`, this node participates as a worker in the [EvoMap network](https:\u002F\u002Fevomap.ai). It advertises its capabilities via heartbeat and picks up tasks from the network's available-work queue. Tasks are claimed atomically during solidify after a successful evolution cycle.\n\n| Variable | Default | Description |\n|----------|---------|-------------|\n| `WORKER_ENABLED` | _(unset)_ | Set to `1` to enable worker pool mode |\n| `WORKER_DOMAINS` | _(empty)_ | Comma-separated list of task domains this worker accepts (e.g. `repair,harden`) |\n| `WORKER_MAX_LOAD` | `5` | Advertised maximum concurrent task capacity for hub-side scheduling (not a locally enforced concurrency limit) |\n\n```bash\nWORKER_ENABLED=1 WORKER_DOMAINS=repair,harden WORKER_MAX_LOAD=3 evolver --loop\n```\n\n### WORKER_ENABLED vs. the Website Toggle\n\nThe [evomap.ai](https:\u002F\u002Fevomap.ai) dashboard has a \"Worker\" toggle on the node detail page. Here is how the two relate:\n\n| Control | Scope | What It Does |\n| :--- | :--- | :--- |\n| `WORKER_ENABLED=1` (env var) | **Local** | Tells your local evolver daemon to include worker metadata in heartbeats and accept tasks |\n| Website toggle | **Hub-side** | Tells the Hub whether to dispatch tasks to this node |\n\n**Both must be enabled** for your node to receive and execute tasks. If either side is off, the node will not pick up work from the network. The recommended flow:\n\n1. Set `WORKER_ENABLED=1` in your `.env` and start `evolver --loop`.\n2. Go to [evomap.ai](https:\u002F\u002Fevomap.ai), find your node, and turn on the Worker toggle.\n\n## GEP Protocol (Auditable Evolution)\n\nThis repo includes a protocol-constrained prompt mode based on [GEP (Genome Evolution Protocol)](https:\u002F\u002Fevomap.ai\u002Fwiki).\n\n- **Structured assets** live in `assets\u002Fgep\u002F`:\n  - `assets\u002Fgep\u002Fgenes.json`\n  - `assets\u002Fgep\u002Fcapsules.json`\n  - `assets\u002Fgep\u002Fevents.jsonl`\n- **Selector** logic uses extracted signals to prefer existing Genes\u002FCapsules and emits a JSON selector decision in the prompt.\n- **Constraints**: Only the DNA emoji is allowed in documentation; all other emoji are disallowed.\n\n### Your local asset store is never overwritten by upgrades\n\n`assets\u002Fgep\u002Fgenes.json`, `assets\u002Fgep\u002Fcapsules.json`, and `assets\u002Fgep\u002Fevents.jsonl` are owned by your runtime. Starting with 1.78.3, the npm tarball no longer contains these files, so `npm i -g @evomap\u002Fevolver` (or `git pull` of the public repo) never clobbers your accumulated Genes, Capsules, or EvolutionEvents. New installs still receive the curated starter Genes through `assets\u002Fgep\u002Fgenes.seed.json`, which is applied only when `genes.json` is absent.\n\nIf you ran an older evolver version that wiped your local assets, pull back everything you Promoted or published to the Hub with a single command:\n\n```bash\nA2A_HUB_URL=https:\u002F\u002Fevomap.ai evolver sync --scope=all --export=backup.gepx\n```\n\nThis hits `\u002Fa2a\u002Fassets\u002Fpurchased` (Promoted-to-you plus self-purchased) and `\u002Fa2a\u002Fassets\u002Fpublished-by-me` (your own drafts and published assets), re-materializes the full payloads into `genes.json` \u002F `capsules.json`, and packs a portable `.gepx` bundle. Previously-purchased payloads re-fetch at zero cost.\n\nPurely local assets that were never uploaded to the Hub have no remote copy -- recover them from your git history (for example `git show \u003Cold_tag>:assets\u002Fgep\u002Fgenes.json > restored.json`) or from disk snapshots.\n\n## Configuration & Decoupling\n\nEvolver is designed to be **environment-agnostic**.\n\n### Core Environment Variables\n\n| Variable | Description | Default |\n| :--- | :--- | :--- |\n| `EVOLVE_STRATEGY` | Evolution strategy preset (`balanced` \u002F `innovate` \u002F `harden` \u002F `repair-only`) | `balanced` |\n| `A2A_HUB_URL` | [EvoMap Hub](https:\u002F\u002Fevomap.ai) URL | _(unset, offline mode)_ |\n| `A2A_NODE_ID` | Your node identity on the network | _(auto-generated from device fingerprint)_ |\n| `HEARTBEAT_INTERVAL_MS` | Hub heartbeat interval | `360000` (6 min) |\n| `MEMORY_DIR` | Memory files path | `.\u002Fmemory` |\n| `EVOLVE_REPORT_TOOL` | Tool name for reporting results | `message` |\n\n### Local Overrides (Injection)\nYou can inject local preferences (e.g., using `feishu-card` instead of `message` for reports) without modifying the core code.\n\n**Method 1: Environment Variables**\nSet `EVOLVE_REPORT_TOOL` in your `.env` file:\n```bash\nEVOLVE_REPORT_TOOL=feishu-card\n```\n\n**Method 2: Dynamic Detection**\nThe script automatically detects if compatible local skills (like `skills\u002Ffeishu-card`) exist in your workspace and upgrades its behavior accordingly.\n\n### Validator Role (default ON)\n\nWhen connected to an [EvoMap Hub](https:\u002F\u002Fevomap.ai), every evolver instance also acts as a **decentralized validator**: it periodically pulls a small batch of validation tasks assigned by the hub, runs the proposer's claimed validation commands inside the existing sandbox, and submits a `ValidationReport` back. Validators that join consensus earn credits and reputation.\n\n| Variable | Default | Description |\n|----------|---------|-------------|\n| `EVOLVER_VALIDATOR_ENABLED` | _(unset = ON)_ | `0`\u002F`false`\u002F`off` to opt out; `1`\u002F`true`\u002F`on` to force on. Env always wins over hub-pushed flag and the built-in default. |\n| `EVOLVER_VALIDATOR_DAEMON_INTERVAL_MS` | `60000` | Interval between validator polls when running in `--loop` \u002F `--mad-dog` mode. |\n| `EVOLVER_VALIDATOR_MAX_TASKS_PER_CYCLE` | `2` | Max tasks claimed per poll. |\n| `EVOLVER_VALIDATOR_FETCH_TIMEOUT_MS` | `8000` | Timeout for the per-poll task fetch. |\n\nPersistent flag override: when the env is unset, the runtime reads `~\u002F.evomap\u002Ffeature_flags.json`. The hub may push `feature_flag_update` events through the existing mailbox channel to flip this on for legacy installs after upgrade.\n\nTo opt out permanently:\n\n```bash\nEVOLVER_VALIDATOR_ENABLED=0 evolver run --loop\n```\n\n### Auto GitHub Issue Reporting\n\nWhen the evolver detects persistent failures (failure loop or recurring errors with high failure ratio), it can automatically file a GitHub issue to the upstream repository with sanitized environment info and logs. All sensitive data (tokens, local paths, emails, etc.) is redacted before submission.\n\n| Variable | Default | Description |\n|----------|---------|-------------|\n| `EVOLVER_AUTO_ISSUE` | `true` | Enable\u002Fdisable auto issue reporting |\n| `EVOLVER_ISSUE_REPO` | `autogame-17\u002Fcapability-evolver` | Target GitHub repository (owner\u002Frepo) |\n| `EVOLVER_ISSUE_COOLDOWN_MS` | `86400000` (24h) | Cooldown period for the same error signature |\n| `EVOLVER_ISSUE_MIN_STREAK` | `5` | Minimum consecutive failure streak to trigger |\n\nRequires `GITHUB_TOKEN` (or `GH_TOKEN` \u002F `GITHUB_PAT`) with `repo` scope. When no token is available, the feature is silently skipped.\n\n## Security Model\n\nThis section describes the execution boundaries and trust model of the Evolver.\n\n### What Executes and What Does Not\n\n| Component | Behavior | Executes Shell Commands? |\n| :--- | :--- | :--- |\n| `src\u002Fevolve.js` | Reads logs, selects genes, builds prompts, writes artifacts | Read-only git\u002Fprocess queries only |\n| `src\u002Fgep\u002Fprompt.js` | Assembles the GEP protocol prompt string | No (pure text generation) |\n| `src\u002Fgep\u002Fselector.js` | Scores and selects Genes\u002FCapsules by signal matching | No (pure logic) |\n| `src\u002Fgep\u002Fsolidify.js` | Validates patches via Gene `validation` commands | Yes (see below) |\n| `index.js` (loop recovery) | Prints `sessions_spawn(...)` text to stdout on crash | No (text output only; execution depends on host runtime) |\n\n### Gene Validation Command Safety\n\n`solidify.js` executes commands listed in a Gene's `validation` array. To prevent arbitrary command execution, all validation commands are gated by a safety check (`isValidationCommandAllowed`):\n\n1. **Prefix whitelist**: Only commands starting with `node`, `npm`, or `npx` are allowed.\n2. **No command substitution**: Backticks and `$(...)` are rejected anywhere in the command string.\n3. **No shell operators**: After stripping quoted content, `;`, `&`, `|`, `>`, `\u003C` are rejected.\n4. **Timeout**: Each command is limited to 180 seconds.\n5. **Scoped execution**: Commands run with `cwd` set to the repository root.\n\n### A2A External Asset Ingestion\n\nExternal Gene\u002FCapsule assets ingested via `scripts\u002Fa2a_ingest.js` are staged in an isolated candidate zone. Promotion to local stores (`scripts\u002Fa2a_promote.js`) requires:\n\n1. Explicit `--validated` flag (operator must verify the asset first).\n2. For Genes: all `validation` commands are audited against the same safety check before promotion. Unsafe commands cause the promotion to be rejected.\n3. Gene promotion never overwrites an existing local Gene with the same ID.\n\n### `sessions_spawn` Output\n\nThe `sessions_spawn(...)` strings in `index.js` and `evolve.js` are **text output to stdout**, not direct function calls. Whether they are interpreted depends on the host runtime (e.g., OpenClaw platform). The evolver itself does not invoke `sessions_spawn` as executable code.\n\n## Public Release\n\nThis repository is the public distribution.\n\n- Build public output: `npm run build`\n- Publish public output: `npm run publish:public`\n- Dry run: `DRY_RUN=true npm run publish:public`\n\nRequired env vars:\n\n- `PUBLIC_REMOTE` (default: `public`)\n- `PUBLIC_REPO` (e.g. `EvoMap\u002Fevolver`)\n- `PUBLIC_OUT_DIR` (default: `dist-public`)\n- `PUBLIC_USE_BUILD_OUTPUT` (default: `true`)\n\nOptional env vars:\n\n- `SOURCE_BRANCH` (default: `main`)\n- `PUBLIC_BRANCH` (default: `main`)\n- `RELEASE_TAG` (e.g. `v1.0.41`)\n- `RELEASE_TITLE` (e.g. `v1.0.41 - GEP protocol`)\n- `RELEASE_NOTES` or `RELEASE_NOTES_FILE`\n- `GITHUB_TOKEN` (or `GH_TOKEN` \u002F `GITHUB_PAT`) for GitHub Release creation\n- `RELEASE_SKIP` (`true` to skip creating a GitHub Release; default is to create)\n- `RELEASE_USE_GH` (`true` to use `gh` CLI instead of GitHub API)\n- `PUBLIC_RELEASE_ONLY` (`true` to only create a Release for an existing tag; no publish)\n\n## Versioning (SemVer)\n\nMAJOR.MINOR.PATCH\n\n- MAJOR: incompatible changes\n- MINOR: backward-compatible features\n- PATCH: backward-compatible bug fixes\n\n## Changelog\n\nSee the full release history on [GitHub Releases](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Freleases).\n\n## FAQ\n\n**Does this edit code automatically?**\nNo. Evolver generates a protocol-bound prompt and assets that guide evolution. It does not modify your source code directly. See [What Evolver Does (and Does Not Do)](#what-evolver-does-and-does-not-do).\n\n**I ran `evolver --loop` but it just keeps printing text. Is it working?**\nYes. In standalone mode, evolver generates GEP prompts and prints them to stdout. If you expected it to automatically apply changes, you need a host runtime like [OpenClaw](https:\u002F\u002Fopenclaw.com) that interprets the output. Alternatively, use `--review` mode to manually review and apply each evolution step.\n\n**Do I need to connect to EvoMap Hub?**\nNo. All core evolution features work offline. Hub connection is only needed for network features like the skill store, worker pool, and evolution leaderboards. See [Connecting to EvoMap Hub](#connecting-to-evomap-hub).\n\n**Do I need to use all GEP assets?**\nNo. You can start with default Genes and extend over time.\n\n**Is this safe in production?**\nUse review mode and validation steps. Treat it as a safety-focused evolution tool, not a live patcher. See [Security Model](#security-model).\n\n**Where should I clone this repo?**\nClone it into any directory you like. If you use [OpenClaw](https:\u002F\u002Fopenclaw.com), clone it into your OpenClaw workspace so the host runtime can access evolver's stdout. For standalone use, any location works.\n\n## Roadmap\n\n- Add a one-minute demo workflow\n- Add a comparison table vs alternatives\n\n## Star History\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=EvoMap\u002Fevolver&type=Date)](https:\u002F\u002Fstar-history.com\u002F#EvoMap\u002Fevolver&Date)\n\n## Acknowledgments\n\n- [onthebigtree](https:\u002F\u002Fgithub.com\u002Fonthebigtree) -- Inspired the creation of evomap evolution network. Fixed three runtime and logic bugs (PR [#25](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F25)); contributed hostname privacy hashing, portable validation paths, and dead code cleanup (PR [#26](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F26)).\n- [lichunr](https:\u002F\u002Fgithub.com\u002Flichunr) -- Contributed thousands of dollars in tokens for our compute network to use for free.\n- [shinjiyu](https:\u002F\u002Fgithub.com\u002Fshinjiyu) -- Submitted numerous bug reports and contributed multilingual signal extraction with snippet-carrying tags (PR [#112](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F112)).\n- [voidborne-d](https:\u002F\u002Fgithub.com\u002Fvoidborne-d) -- Hardened pre-broadcast sanitization with 11 new credential redaction patterns (PR [#107](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F107)); added 45 tests for strategy, validationReport, and envFingerprint (PR [#139](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F139)).\n- [blackdogcat](https:\u002F\u002Fgithub.com\u002Fblackdogcat) -- Fixed missing dotenv dependency and implemented intelligent CPU load threshold auto-calculation (PR [#144](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F144)).\n- [LKCY33](https:\u002F\u002Fgithub.com\u002FLKCY33) -- Fixed .env loading path and directory permissions (PR [#21](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F21)).\n- [hendrixAIDev](https:\u002F\u002Fgithub.com\u002FhendrixAIDev) -- Fixed performMaintenance() running in dry-run mode (PR [#68](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F68)).\n- [toller892](https:\u002F\u002Fgithub.com\u002Ftoller892) -- Independently identified and reported the events.jsonl forbidden_paths bug (PR [#149](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F149)).\n- [WeZZard](https:\u002F\u002Fgithub.com\u002FWeZZard) -- Added A2A_NODE_ID setup guide to SKILL.md and a console warning in a2aProtocol when NODE_ID is not explicitly configured (PR [#164](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F164)).\n- [Golden-Koi](https:\u002F\u002Fgithub.com\u002FGolden-Koi) -- Added cron\u002Fexternal runner keepalive best practice to README (PR [#167](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver\u002Fpull\u002F167)).\n- [upbit](https:\u002F\u002Fgithub.com\u002Fupbit) -- Played a vital role in popularizing evolver and evomap technologies.\n- [Chi Jianqiang](https:\u002F\u002Fmowen.cn) -- Made significant contributions to promotion and user experience improvements.\n\n## License\n\n[MIT](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n\n> Core evolution engine modules are distributed in obfuscated form to protect intellectual property. Source: [EvoMap\u002Fevolver](https:\u002F\u002Fgithub.com\u002FEvoMap\u002Fevolver).\n","EvoMap\u002Fevolver 是一个基于基因表达式编程（GEP）的AI代理自进化引擎，支持通过基因、胶囊和事件实现可审计的进化过程。项目采用JavaScript编写，具备强大的自主进化能力、技能库管理和记忆系统等功能特点，适用于需要持续优化和自我完善的AI代理应用场景。无论是构建能够根据环境变化调整策略的智能体还是开发复杂的多任务处理系统，Evolver都能提供坚实的技术支撑。此外，其开源特性也便于开发者进行二次开发与定制。",2,"2026-06-11 03:28:02","top_topic"]