Skillify transforms a proven, repeatable workflow into a concise, portable, and behaviorally evaluated Agent Skill.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skillify
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install skillify using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skillify?
Skillify is a meta-skill from the Overpowered suite for productizing workflows that have already succeeded in real use. It extracts the invariant method, important corrections, required tools, gotchas, stopping conditions, and output contract while removing accidental project context.
Designed for Openclaw Skills and other portable Agent Skills environments, Skillify emphasizes precise trigger metadata, minimal instructions, progressive disclosure through references, and realistic behavioral evaluations. It is not a conversation-saving tool or a mechanism for inventing capabilities during an active task.
Skillify Use Cases
- Graduate a qualified Skill Academy capability into a durable Agent Skill.
- Package a workflow that has succeeded repeatedly and appears reusable across projects.
- Convert execution evidence into concise instructions, trigger metadata, and an explicit output contract.
- Capture important corrections, non-obvious gotchas, required tools, and reliable stopping conditions.
- Create regression evals for failures discovered during real execution.
- Validate whether a new Openclaw Skills package adds value compared with a no-skill or previous-version baseline.
- Draft an explicitly experimental skill when evidence is incomplete, without incorrectly labeling it proven or graduated.
How Skillify Works
- Inspect execution history or Skill Academy evidence and identify what actually created value.
- Extract the reusable sequence, meaningful corrections, non-obvious gotchas, required capabilities, stopping condition, and output contract.
- Remove project-specific names, paths, values, and assumptions unless they intentionally define the target domain.
- Define one coherent responsibility and split the package only when its parts can stand alone.
- Write a precise
descriptioncovering both the skill capability and the conditions that should trigger it. - Keep
SKILL.mdconcise and move deeper material into one-levelreferences/files when necessary. - Add two or three realistic evals, including a boundary case that should not trigger and regression cases for observed failures.
- Validate Agent Skills structure, naming, metadata, and package organization.
- Test from a clean context against a no-skill or prior-version baseline when the host environment permits.
- Remove instructions that do not measurably improve behavior.
- Record the deployment target and version only after the human or project graduation gate is satisfied.
- Stop when the package is structurally valid, scoped coherently, trigger metadata is precise, evals distinguish the desired behavior from the baseline, and deployment status is stated truthfully.
Skillify Setup
Skillify is a packaging workflow rather than a conventional runtime dependency. Start with real execution evidence or a capability marked qualified by the project’s Skill Academy policy.
- Create the portable skill directory and expected package structure:
mkdir -p skill-name/references skill-name/evals
touch skill-name/SKILL.md skill-name/evals/evals.json
- Populate
SKILL.mdwith concise frontmatter, a precise description, the invariant procedure, trigger conditions, gotchas, and truthful maturity status. - Add reference files only when detailed material cannot remain concise; keep references one level below the skill directory.
- Add two or three realistic behavioral evaluations to
evals/evals.json, including non-trigger and regression scenarios. - Validate the package with the host Agent Skills harness or external eval runner. Behavioral evaluation execution depends on the environment.
- Compare clean-context behavior with a no-skill or prior-version baseline where supported.
- Do not install the result globally by default. Deployment and graduation remain separate authority decisions.
The package should ultimately resemble:
skill-name/
├── SKILL.md
├── references/ # only when needed
└── evals/
└── evals.json
Skillify Data Schema & Taxonomy
Skillify organizes each deliverable as a small portable Agent Skill package:
| Path or field | Purpose |
|---|---|
SKILL.md |
Primary skill instructions, trigger metadata, scope, workflow, gotchas, and output contract. |
references/ |
Optional one-level directory for deeper procedural details and supporting material. |
evals/evals.json |
Behavioral evaluations used to test desired behavior, boundary handling, and regressions. |
name |
Stable skill identifier using valid Agent Skills naming conventions. |
description |
Trigger-focused metadata describing both what the skill does and when it should activate. |
| Evidence status | Indicates whether the package is qualified, experimental, tested, graduated, or deployed; claims must match actual evidence. |
| Version and deployment target | Recorded only after the relevant human or project graduation gate is satisfied. |
The conceptual taxonomy includes reusable sequence, corrections, gotchas, required capabilities, stopping condition, and output contract. Project-specific values should be removed unless they intentionally define a domain-specific skill. A conversation summary, one-off successful answer, or generic advice does not qualify as the skill’s data model.
Skillify Advanced Features
- Evidence-gated graduation prevents unproven workflows from being presented as production-ready Openclaw Skills.
- Trigger-focused descriptions improve activation precision and reduce accidental over-triggering.
- Progressive disclosure keeps
SKILL.mdminimal while allowing optional reference documents. - Boundary evaluations test when the skill should not activate, not just successful execution paths.
- Regression evaluations preserve fixes for observed failures and corrections.
- Clean-context and baseline comparisons help measure whether the skill adds value.
- Portable packaging avoids unnecessary dependence on runtime-specific tool names.
- Explicit stopping conditions prevent endless refinement once structure, scope, triggers, and eval coverage are sufficient.
- Separation of packaging from deployment preserves human or project authority over global installation and graduation.
- Experimental drafts can be created transparently when evidence is incomplete, without making unsupported claims of generality or testing.
SKILL.md
Loading
Related Openclaw Skills
Discovering Viral TikTok Content Trends
This skill analyzes high-performing TikTok posts to uncover repeatable viral formats, hooks, engagement patterns, and content opportunities for a specific niche.
Discovering TikTok Creators by Niche
Discover, enrich, filter, and rank TikTok creators for influencer campaigns using niche hashtags and Apify scrapers.
Discovering Brand Ambassadors Across Platforms
Discover, classify, and rank potential brand ambassadors across Instagram, TikTok, and Twitter using Apify-powered social data collection.
Analyzing Competitor YouTube Content Strategy
This skill reverse-engineers a competitor’s YouTube strategy to reveal the topics, formats, publishing patterns, and performance signals that drive results.
CONNSKILL Growth Services
CONNSKILL Growth Services gives agents pay-per-call access to SEO intelligence, SERP data, verification services, EU-hosted AI and x402 seller trust checks using USDC on Base.
WeChat Channels Ad Audit
A structured WeChat Channels creator audit skill that turns account, content, and engagement data into investment-ready advertising reports.
Featured*
