← Back to Skills Marketplace
Skill Autogenesis
by
codeblackhole
· GitHub ↗
· v1.3.2
· MIT-0
120
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install skill-autogenesis
Description
Review completed work, summarize reusable procedures, identify recurring workflow patterns, and decide whether to create a skill, patch an existing skill, st...
Usage Guidance
This skill is internally consistent and intentionally designed to convert recurring, verified workflows into skill files, but it will actively evaluate many completed tasks if loaded early in a session. Before installing or enabling it: (1) confirm you trust agents with the skill_manage tool and their ability to write to your skills directory (e.g., ~/.hermes/skills/); (2) decide whether you want autonomous creation/patching or prefer to require human approval — monitor outputs for the required internal classification template and recommended_action before any file writes; (3) ensure no secrets or sensitive information can be accidentally captured in generated skill content; (4) consider running it in a non-sensitive test session first so you can observe how often it proposes writes; (5) keep backups of your skills dir or review created files regularly and restrict agent tool permissions if you want tighter control.
Capability Analysis
Type: OpenClaw Skill
Name: skill-autogenesis
Version: 1.3.2
The skill-autogenesis bundle is a meta-utility designed to help AI agents convert successful task workflows into reusable procedural memory (skills). It provides a structured framework for the agent to analyze its own actions, detect recurrence, and decide whether to create or patch skill files using tools like `skill_manage`. The bundle is characterized by extensive defensive logic, including a 'Decision Matrix' and 'Hard Stop Rules' (found in SKILL.md and references/hard-stop-rules.md) specifically designed to prevent the accidental creation of skills from sensitive data, user preferences, or one-off results. There is no evidence of malicious intent, data exfiltration, or unauthorized execution; the logic is entirely focused on organized knowledge management within the agent's designated environment.
Capability Tags
Capability Assessment
Purpose & Capability
The name and description match the actual behavior: distill workflows and, when appropriate, create/patch skills or store memory. It does not request unrelated binaries, environment variables, or network endpoints. All referenced capabilities (session_search, memory, skill_manage) are reasonable for the stated goal.
Instruction Scope
SKILL.md instructs the agent to inspect completed tasks, produce a classification template, estimate recurrence/stability, and only then call skill_manage(create|patch) or write memory. This stays within the skill's purpose. Note: the skill is designed to evaluate every 'meaningful' success by default, so if loaded early it will observe/consider many tasks — the gating (classification template and hard-stop rules) is explicit but users should be aware the agent will routinely evaluate workflows while the skill is active.
Install Mechanism
Instruction-only skill with no install spec and no code files. No downloads, binaries, or package installs are required, which minimizes persistence and supply-chain risk.
Credentials
The package declares no required environment variables, credentials, or config paths. The instructions reference tool APIs (skill_manage, session_search, memory) but do not require unrelated secrets or external credentials.
Persistence & Privilege
The skill can recommend or call skill_manage to create/patch skill files when the skill_manage tool is available — this is coherent with its purpose. It does not force always:true. Users should be aware that, if the agent has access to skill-management tools and file-write permissions (typical for agent-managed skills), the agent may autonomously write files into user skill directories (e.g., ~/.hermes/skills/) according to the documented gating rules.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install skill-autogenesis - After installation, invoke the skill by name or use
/skill-autogenesis - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.2
# Changelog
## 1.3.2
- Added hard stop rules that explicitly block invalid `skill_manage(create)` usage.
- Tightened lifecycle rules so only `knowledge_type=procedure` may enter procedural skill actions.
- Added `references/hard-stop-rules.md` with a fast test for valid skill candidates.
- Updated the README to surface hard interception behavior.
## 1.3.1
- Added a mandatory internal classification template before any persistence action.
- Added a checklist gate that must fully pass before create or patch is allowed.
- Extended verification rules so create and patch require a completed decision record.
- Updated the README to emphasize classification template first, file writes second.
## 1.3.0
- Added an explicit decision matrix to route outcomes to skill, memory, prompt, or no-op.
- Added an output contract so agents classify the result before writing files.
- Added a recurrence warning that repeated rules still do not qualify as skills without executable procedure.
- Added `references/classification-examples.md` with concrete examples of what should and should not become a skill.
## 1.2.2
- Changed the default posture from implied creation to explicit classification first.
- Added a hard gate that only executable reusable procedures may become skills.
- Added routing guidance for non-skill outcomes such as preferences, policies, boundaries, and prompt-level governance.
- Added a pitfall warning that rules are not skills unless they contain trigger, action, and verification.
- Updated the bilingual README to reduce the common misread that this skill should create a new rule after every success.
## 1.2.1
- Refined positioning to emphasize verification-gated skill creation rather than unconditional autonomous writes.
- Clarified that skill creation happens only when recurrence, stability, and environment policy permit it.
- Tightened README wording to better communicate lifecycle controls and safety boundaries.
- Kept local fallback reference behavior for GitHub sources.
## 1.2.0
- Added explicit source resolution policy: GitHub first, local fallback second, `[UNVERIFIED]` last.
- Added local fallback reference files under `references/fallback/`.
- Added `skill_manage`-style lifecycle handling for create, patch, edit, write_file, remove_file, and guarded delete.
- Added reusable template file for generated skills.
## 1.1.0
- Added lifecycle-oriented behavior modeled after `skill_manage`.
- Added support for supporting-file management and duplicate-skill avoidance.
## 1.0.0
- Initial release.
- Added automatic workflow distillation, recurrence detection, and skill generation guidance.
v1.2.1
Clarified lifecycle safeguards, added release changelog, and emphasized verification-gated skill creation.
v1.2.0
Initial public release of skill-autogenesis with automatic skill distillation, skill_manage-style lifecycle handling, and local fallback references for GitHub sources.
Metadata
Frequently Asked Questions
What is Skill Autogenesis?
Review completed work, summarize reusable procedures, identify recurring workflow patterns, and decide whether to create a skill, patch an existing skill, st... It is an AI Agent Skill for Claude Code / OpenClaw, with 120 downloads so far.
How do I install Skill Autogenesis?
Run "/install skill-autogenesis" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Skill Autogenesis free?
Yes, Skill Autogenesis is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Skill Autogenesis support?
Skill Autogenesis is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Skill Autogenesis?
It is built and maintained by codeblackhole (@codeblackhole1024); the current version is v1.3.2.
More Skills