← Back to Skills Marketplace
OpenAuto - Autonomous Agent
by
danavfrost
· GitHub ↗
· v1.0.7
· MIT-0
130
Downloads
1
Stars
0
Active Installs
8
Versions
Install in OpenClaw
/install openauto
Description
Proactive, self-improving agent architecture with hardened security. WAL Protocol, Working Buffer, Compaction Recovery, Autonomous Crons, and battle-tested p...
Usage Guidance
This skill appears to be what it says: a proactive agent framework that stores and manages local workspace files and includes a benign security-audit script. Before installing: (1) don't place secrets (API keys, passwords) in USER.md, memory files, or other workspace files the agent will read; keep secrets in a properly permissioned .credentials/ directory and verify .gitignore. (2) Review assets/AGENTS.md, TOOLS.md and the security-audit.sh script yourself and only run the audit when you trust the environment (it will read ~/.openclaw/openclaw.json if present). (3) Remember the documented guardrails are advisory — the agent's compliance depends on the model; test in isolation and avoid storing highly sensitive data in the workspace until you're confident in behavior.
Capability Assessment
Purpose & Capability
The name/description (proactive, self-improving agent) matches what the skill does: it reads/writes agent identity, user context, memory, onboarding, tools, and heartbeat files and provides guidance for proactive behavior. No unrelated binaries or external credentials are requested.
Instruction Scope
SKILL.md explicitly lists the workspace files the agent will read/write and limits automatic startup reads to SOUL.md, USER.md, and today's daily notes. That scope aligns with a proactive agent, but the guardrails are advisory (the doc notes they are not technically enforced), so correct behavior depends on the model following instructions. The skill also documents that .credentials/ and ~/.openclaw/openclaw.json are read only on request (e.g., when running the audit script).
Install Mechanism
No install spec is present; the skill is instruction-first with a small included audit script. No downloads or external installers are used, which is the lowest-risk install profile.
Credentials
The skill declares no required environment variables or primary credential. Files it reads/writes are workspace-local and documented. The only non-workspace path referenced (~/.openclaw/openclaw.json) is accessed only when running security-audit.sh, which is consistent with its stated auditing purpose.
Persistence & Privilege
The skill is not force-installed (always: false). It does not request persistent platform privileges or attempt to modify other skills or system-wide settings. Autonomous invocation is allowed by default but not combined with other concerning privileges.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openauto - After installation, invoke the skill by name or use
/openauto - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.7
## OpenAuto 1.0.7 Changelog
- Improved filesystem and credential access documentation for clarity and security transparency.
- Clarified that `.credentials/` and `openclaw.json` are never read automatically at startup—access is only on explicit user request.
- Added emphasis on which files are read during automatic startup.
- Updated references to OpenAuto v1.0.7 throughout the documentation.
- No functional changes to code or behavior; documentation and trust model improvements only.
v1.0.6
- Updated filesystem access documentation: added `.credentials/` directory as the expected location for credentials, clarifying that it is user-managed and not written to by the skill.
- Expanded notes on trust and skill metadata: explained that the OpenClaw metadata spec does not support declaring filesystem paths, using a table as a disclosure substitute.
- No functional code or logic changes in this release; documentation only.
- Lowered version to 1.0.6 to reflect alignment with OpenClaw conventions.
v1.0.5
- Corrected documentation.
v1.0.4
openauto v1.1.0
- Added a clear trust model to the documentation, clarifying advisory (not enforced) nature of agent safety rules regarding file access and secret handling.
- Updated documentation to include guidance about not storing credentials or highly sensitive data in workspace files accessible to the agent.
- No functional changes to code or file access; documentation improvements only.
- Added support for OpenMem memory management.
v1.0.3
Documentation updated to reflect operations.
v1.0.2
Changed loader wording.
v1.0.1
Corrected the anti-injection instructions to be more clear without using injection phrases.
v1.0.0
openauto v1.0.0 — Initial release featuring a security-hardened, self-improving agent architecture.
- Forked from proactive-agent v3.1.0 with a thorough security audit: removed unsafe principles, fixed config paths, scrubbed fabricated statistics, and introduced privacy guardrails for Working Buffer.
- Implements WAL Protocol for robust correction/decision logging.
- Adds Working Buffer and Compaction Recovery features for improved persistence across memory flushes and context truncation.
- Introduces Autonomous Crons, reverse prompting, and enhanced agent self-improvement systems.
- Comprehensive documentation included for architecture, protocols, quick start, and security features.
Metadata
Frequently Asked Questions
What is OpenAuto - Autonomous Agent?
Proactive, self-improving agent architecture with hardened security. WAL Protocol, Working Buffer, Compaction Recovery, Autonomous Crons, and battle-tested p... It is an AI Agent Skill for Claude Code / OpenClaw, with 130 downloads so far.
How do I install OpenAuto - Autonomous Agent?
Run "/install openauto" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is OpenAuto - Autonomous Agent free?
Yes, OpenAuto - Autonomous Agent is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does OpenAuto - Autonomous Agent support?
OpenAuto - Autonomous Agent is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created OpenAuto - Autonomous Agent?
It is built and maintained by danavfrost (@danavfrost); the current version is v1.0.7.
More Skills