How to use Claude to create reusable skills
The author reports that Wiz created and used a Typefully skill, gained a scheduled error-resolution loop, and moved job-board work to parallel sub-agents.
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
For [requested recurring task], inspect [existing skills directory] for coverage. If a reusable capability is missing, research [service options], create a skill with API documentation and a clear description, use it for the requested task, and explain what you added.
Reach people exploring AI agents
See placements →What happened
The author reports that Wiz created and used a Typefully skill, gained a scheduled error-resolution loop, and moved job-board work to parallel sub-agents.
Reported by joozio on dev.toHow the agent did it
- Migrate existing capabilities into native Claude Code skills.
- Add automatic skill-creation criteria to CLAUDE.md.
- Identify a missing social-posting capability.
Where a person stepped in
The author configured and rebuilt the system; the example ends with Wiz informing the author that it created a skill. No approval step is described.
A person set the goal and reviewed the result.
The author spent a week rebuilding Wiz, a personal agent powered by Claude Code. The earlier version relied on custom skill folders, complicated memory infrastructure, and sequential work. The rebuild moved capabilities into Claude Code's native skills directory and workers into its agents directory, with YAML frontmatter describing their roles.
Thirteen existing skills were migrated. The author also added instructions in CLAUDE.md allowing automatic skill creation when a request exposes a reusable, structured capability gap that benefits future work. In the reported social-posting example, Wiz recognized a missing capability, researched direct API and Typefully options, created a Typefully skill with API documentation, used it immediately, and informed the author.
The rebuild also introduced an error registry recording context, severity, solutions, and resolution status. A scheduled morning task reads the registry and attempts to fix high-severity unresolved errors. For job-board research, the author configured Haiku sub-agents to scrape sources simultaneously and aggregate their results.
The reported result is an agent that can develop capabilities, record and address errors, and execute tasks in parallel. Who it's for: Claude Code users with repeated personal automation tasks who want reusable skills rather than rebuilding instructions each session.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: dev.to · joozio
Frequently asked questions
When does Wiz create a new skill?
The instructions require a capability gap, likely reuse, a structured process or tool interaction, and a benefit for future user requests.
How does the error registry get used?
A task scheduled for 7:15 AM reads the registry, identifies high-severity unfixed errors, attempts resolution, and marks resolved entries.
More details
Remaining steps
- Research options and create the Typefully skill.
- Use the new skill and report its creation.
About this account
Wiz creates reusable skills when explicit criteria are met and uses the resulting capability for the requested task.
Platform: dev.to. Attribution: https://dev.to/joozio/when-your-ai-agent-starts-fixing-itself-a-week-of-rebuilding-wiz-31an. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: joozio on dev.to
Source published 7 Mar 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Typefully · CLAUDE.md
Skills · Automation · Error recovery
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.