Description

A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution.

README (SKILL.md)

🧬 Capability Evolver

Name: Evolver.Bak
Author: chris8265-cl

"Evolution is not optional. Adapt or die."

The Capability Evolver is a meta-skill that allows OpenClaw agents to inspect their own runtime history, identify failures or inefficiencies, and autonomously write new code or update their own memory to improve performance.

Features

Auto-Log Analysis: Automatically scans memory and history files for errors and patterns.
Self-Repair: Detects crashes and suggests patches.
GEP Protocol: Standardized evolution with reusable assets.
One-Command Evolution: Just run /evolve (or node index.js).

Usage

Standard Run (Automated)

Runs the evolution cycle. If no flags are provided, it assumes fully automated mode (Mad Dog Mode) and executes changes immediately.

node index.js

Review Mode (Human-in-the-Loop)

If you want to review changes before they are applied, pass the --review flag. The agent will pause and ask for confirmation.

node index.js --review

Mad Dog Mode (Continuous Loop)

To run in an infinite loop (e.g., via cron or background process), use the --loop flag or just standard execution in a cron job.

node index.js --loop

Configuration

Environment Variable	Default	Description
`EVOLVE_ALLOW_SELF_MODIFY`	`false`	Allow evolution to modify evolver's own source code. NOT recommended for production. Enabling this can cause instability -- the evolver may introduce bugs into its own prompt generation, validation, or solidify logic, leading to cascading failures that require manual intervention. Only enable for controlled experiments.
`EVOLVE_LOAD_MAX`	`2.0`	Maximum 1-minute load average before evolver backs off.
`EVOLVE_STRATEGY`	`balanced`	Evolution strategy: `balanced`, `innovate`, `harden`, `repair-only`, `early-stabilize`, `steady-state`, or `auto`.

GEP Protocol (Auditable Evolution)

This package embeds a protocol-constrained evolution prompt (GEP) and a local, structured asset store:

assets/gep/genes.json: reusable Gene definitions
assets/gep/capsules.json: success capsules to avoid repeating reasoning
assets/gep/events.jsonl: append-only evolution events (tree-like via parent id)

Emoji Policy

Only the DNA emoji is allowed in documentation. All other emoji are disallowed.

Configuration & Decoupling

This skill is designed to be environment-agnostic. It uses standard OpenClaw tools by default.

Local Overrides (Injection)

You can inject local preferences (e.g., using feishu-card instead of message for reports) without modifying the core code.

Method 1: Environment Variables Set EVOLVE_REPORT_TOOL in your .env file:

EVOLVE_REPORT_TOOL=feishu-card

Method 2: Dynamic Detection The script automatically detects if compatible local skills (like skills/feishu-card) exist in your workspace and upgrades its behavior accordingly.

Safety & Risk Protocol

1. Identity & Directives

Identity Injection: "You are a Recursive Self-Improving System."
Mutation Directive:
- If Errors Found -> Repair Mode (Fix bugs).
- If Stable -> Forced Optimization (Refactor/Innovate).

2. Risk Mitigation

Infinite Recursion: Strict single-process logic.
Review Mode: Use --review for sensitive environments.
Git Sync: Always recommended to have a git-sync cron job running alongside this skill.

Before Troubleshooting -- Check Your Version First

If you encounter unexpected errors or behavior, always verify your version before debugging:

node -e "const p=require('./package.json'); console.log(p.version)"

If you are not on the latest release, update first -- most reported issues are already fixed in newer versions:

# If installed via git
git pull && npm install

# If installed via npm (global install)
npm install -g evolver@latest

Latest releases and changelog: https://github.com/autogame-17/evolver/releases

License

MIT

Usage Guidance

Things to check before installing or running this skill: - Do not run it in your production workspace without review. Run it first in a disposable/sandbox git repo. - Ensure Git is available. The code expects git but the registry metadata did not declare it. - Protect secrets: the process loads .env on startup. Remove or move any sensitive keys (GITHUB_TOKEN, GH_TOKEN, other PATs) from any repo .env you run this in, or run with an empty .env. Assume any node script run by the skill can read environment variables. - Prefer review mode: run node index.js --review (or run single-run mode) rather than automatic loop mode. Keep EVOLVE_ALLOW_SELF_MODIFY=false (default) unless you explicitly want the engine to edit its own source and you have manual rollback procedures. - Inspect network code: before enabling any publish/persist/emit flags (A2A, a2a_export with --persist, A2A_EMIT_DECISIONS, or publishing scripts), open src/gep/a2aProtocol (and related transport code) to confirm what endpoints and auth it uses. - Audit validators: solidify will execute validation commands but restricts them to node/npm/npx. Still, review any Gene validation scripts or validation commands to ensure they don't perform unwanted actions (network calls, credential upload, etc.). - If you accept it, run it with least privilege: in a cloned test repo without remote credentials, and monitor what files it changes (use git to review diffs). If you plan to allow any automated promotions (a2a_promote) require that operators perform local manual verification first. Summary recommendation: the project appears functionally consistent with a self-evolution tool, but because important environment assumptions are undeclared and the runtime can read .env, run node scripts, and perform network A2A operations, treat it as potentially risky until you inspect transports, validators, and run it in a sandboxed environment.

Capability Analysis

Type: OpenClaw Skill Name: evolver-bak Version: 1.0.0 The 'Capability Evolver' is a meta-skill designed for autonomous agent self-improvement through code modification and execution. It possesses high-risk capabilities, including the ability to execute shell commands for validation (src/gep/solidify.js), communicate with an external hub at evomap.ai (src/gep/a2aProtocol.js), and run in an infinite loop (index.js). While the bundle includes safety mechanisms such as credential sanitization (src/gep/sanitize.js), command whitelisting, and path protection, the core logic allows for the ingestion and execution of 'genes' (strategies) which could lead to unauthorized actions if safety checks are bypassed. The use of sessions_spawn to trigger recursive agent cycles and the 'Mad Dog Mode' for persistence make this bundle highly sensitive and potentially dangerous in unconstrained environments.

Capability Tags

cryptorequires-walletcan-make-purchasesrequires-sensitive-credentials

Capability Assessment

⚠ Purpose & Capability

The skill's code matches the stated purpose (log analysis, Genes/Capsules, GEP/solidify, self-repair), however the registry metadata claims no required binaries while the code and README require Git (execSync('git ...') is used in build_public and other modules). That undeclared dependency is an incoherence. Other optional capabilities (publishing releases, A2A transport) are present in code but not surfaced as required permissions/credentials in the registry metadata.

⚠ Instruction Scope

SKILL.md and the code instruct running node index.js (or loop/solidify). The runtime reads .env early, scans repository files, reads/writes assets under assets/ and memory/, uses git operations, and will execute validation commands during solidify. Those actions go beyond simple 'analysis' and include repository modifications and running local commands. The README/safety text describes protections (review mode, validation gate) but the instructions allow fully automated 'Mad Dog / loop' mode that makes file changes and can restart itself—so the runtime scope is broad and potentially impactful.

✓ Install Mechanism

No external download/install spec is declared (no remote installers, no URL extracts). The repo includes many source files and a small npm dependency (dotenv). Since nothing is fetched from arbitrary URLs at install time, install risk is lower. However, the package will run local filesystem and child processes when executed.

⚠ Credentials

The registry lists no required env vars, but the code loads .env (dotenv) at startup and supports many environment flags (EVOLVE_ALLOW_SELF_MODIFY, EVOLVE_REPORT_TOOL, EVOLVE_STRATEGY, GITHUB_TOKEN/GH_TOKEN noted in README, A2A_* variables, etc.). The loader reads .env before anything else, which means any secrets in .env are accessible to the process. Solidify runs validation commands (allowed prefixes restricted to node/npm/npx) — but arbitrary node scripts can still access env and network. A2A export/ingest/publish code paths call getTransport().send(), indicating potential network I/O for asset sharing. These env/credential uses are not fully declared in metadata and could be disproportionate if you expected a read-only analyzer.

ℹ Persistence & Privilege

The skill is not marked always:true, but it supports an internal daemon (--loop) that writes an evolver.pid, can spawn a child to restart itself, and writes artifacts to assets/ and memory/. It can therefore persist state and run continuously if launched. Self-modification is gated by EVOLVE_ALLOW_SELF_MODIFY (default false) and review mode is available; nonetheless, combined with autonomous invocation and loop behavior this increases blast radius if enabled or misconfigured.

Version History

v1.0.0

Initial release of capability-evolver. - Introduces automated self-evolution for AI agents, including auto-log analysis and self-repair routines. - Supports multiple evolution strategies and modes (automated, review, infinite loop). - Implements auditable GEP protocol with structured local asset storage. - Provides robust configuration via environment variables and dynamic local overrides. - Includes risk mitigation features like single-process enforcement and git sync recommendation. - Documentation enforces DNA emoji only, supporting clear, focused communication.

Metadata

Slug evolver-bak

Version 1.0.0

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 1

Frequently Asked Questions

What is Evolver.Bak?

A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution. It is an AI Agent Skill for Claude Code / OpenClaw, with 80 downloads so far.

How do I install Evolver.Bak?

Run "/install evolver-bak" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Evolver.Bak free?

Yes, Evolver.Bak is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Evolver.Bak support?

Evolver.Bak is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Evolver.Bak?

It is built and maintained by chris8265-cl (@chris8265-cl); the current version is v1.0.0.

More Skills

Evolver.Bak