Skill Self-Improvement
Skill authoring skill, available on Zeplik
Skill Self-Improvement is a ready-to-run skill authoring skill on Zeplik. Capture a hard-won, reusable procedure from the current conversation as a personal skill via skill_manage, so the next occurrence is solved in one step. Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.
The Skill Self-Improvement skill loads automatically when your request matches it, or you can invoke it directly by typing /self-improve in any chat. It works with attachments, connectors, and any model that supports the task, so you get the same expert method every time without setting anything up.
What the Skill Self-Improvement skill can do
- Detect when a hard-won multi-step fix is worth saving as a skill
- Ask permission before capturing anything into the catalog
- Check existing skills to update instead of creating duplicates
- Distill the working procedure into a named, versioned skill file
Try these prompts on Zeplik
Pick a prompt to open it in the Zeplik app. If you are not signed in yet, your prompt is waiting for you the moment you do.
How the Skill Self-Improvement skill works
Self improve
When a conversation produces knowledge that took real effort to earn and will be needed again, offer to preserve it as a personal skill. The goal is that the next time this problem appears, it is solved in one step instead of ten.
When to offer
Offer capture when all of these hold:
- Non-obvious: the solution required iteration, a misleading error message, a workaround, or trial and error. Not something one search would answer.
- Multi-step: there is a procedure worth writing down, not a single fact.
- Recurring: the user will plausibly face this again, in this project or a similar one.
Also respond when the user asks directly: "save this as a skill", "remember this", "what did we learn?".
Do not offer for routine tasks, single lookups, guesses that were never confirmed to work, or knowledge already covered well by official docs. Most conversations should end without an offer; over-capture buries the good skills under noise.
Steps
- Ask first. One sentence: name the procedure you would capture and why it seems worth keeping. Create nothing until the user agrees.
- Check for an existing home. Call skills_list and, for anything adjacent, skill_view. Same trigger and same fix: update that skill instead of creating a duplicate. Same domain but a genuinely different problem: a new skill is fine. Existing skill turned out wrong: update it to the corrected procedure.
- Distill the procedure. Identify the trigger (the exact symptom, error text, or situation), the steps that actually worked in their final form (not the detours), and how to verify success. Strip conversation-specific details that will not generalize; keep specifics like exact error messages that make the skill findable.
- Draft name, description, and body per the guidance below and show the draft in chat.
- On approval, persist with skill_manage: action "create" for a new skill, or action "update" with a bumped version for an existing one. Tell the user it is saved and will route automatically or via "/" plus its name.
Shape of a captured skill
- Name: short kebab-case describing the problem domain, such as "csv-encoding-fix", not "the-thing-from-tuesday".
- Description: third person, and specific about the trigger. Include the concrete symptoms or error text, because that is what routing matches against. "Fix garbled characters when a CSV exported from Excel is loaded; use when accented characters appear as mojibake" will fire when needed; "helps with CSV problems" will not.
- Body: small and procedural. Problem in one or two lines, trigger conditions, numbered solution steps, how to verify it worked, and any caveat discovered the hard way. A few hundred to two thousand characters. Link to official docs for background rather than restating them.
Rules
- Only capture what was verified to work in this conversation. A plausible but untested idea is not a skill.
- Generalizable knowledge only. "This user's API key is X" is never a skill; "this API returns 200 with an error body, check the status field" is.
- Prefer updating over creating. Two skills with overlapping triggers make routing worse for both.
- No secrets, tokens, or private URLs in the captured body.
- The user owns the catalog. If they decline, drop it without argument.
How to use the Skill Self-Improvement skill
Sign in to Zeplik
Create a free Zeplik account or sign in. New accounts start with free credits, so you can try the Skill Self-Improvement skill right away.
Describe your skill authoring task
Ask in plain language, or type /self-improve to invoke the skill directly. Zeplik recognizes the Skill Self-Improvement skill and applies its method.
Review and refine the result
Zeplik returns a clear, structured answer. Ask follow-ups in the same chat to refine it or take the next step.
Source and credit
- Author
- Siqi Chen
- License
- MIT
Adapted from the open-source blader/Claudeception project and tuned to run natively on Zeplik. View source on GitHub.
Frequently asked questions
- What is the Skill Self-Improvement skill?
- Skill Self-Improvement is a ready-to-run skill authoring skill on Zeplik. Capture a hard-won, reusable procedure from the current conversation as a personal skill via skill_manage, so the next occurrence is solved in one step. Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.
- How do I use Skill Self-Improvement on Zeplik?
- Sign in to Zeplik and ask in plain language, or type /self-improve in any chat to invoke it directly. The skill applies its method and returns a result you can refine in the same conversation.
- Which AI model does the Skill Self-Improvement skill use?
- Any model you choose. Zeplik works across every model in one chat, so the Skill Self-Improvement skill runs on your preferred model for the task.
- Where does the Skill Self-Improvement skill come from?
- The Skill Self-Improvement skill is adapted from the open-source blader/Claudeception project (MIT) and tuned to run natively on Zeplik. The original source is linked on this page.
- How much does the Skill Self-Improvement skill cost?
- Using the skill is free to start. You only spend Zeplik credits when the assistant runs, and new accounts begin with free credits.
Related skill authoring skills
- Claude Code ExtensionsAuthoring Claude Code extensions — subagents, hooks, slash commands, MCP server config, output styles, status lines, and skill authoring. Use for "build/configure a Claude Code hook/subagent/command"; for using Claude Code itself see the claude-code guide.
- Evaluation MethodologyUse when designing eval suites for agent skills or plugins -- fixtures, graders, pass bars, dimensions. Not for judging one skill draft (use skill-judge) or LLM app evals (use llm-evaluation).
- Skill CreatorGuide the user through creating a personal skill: interview briefly, draft the name, description, and body, get approval, then persist it with the skill_manage tool. Use when the user wants to create, design, or improve a custom skill, or says things like "make me a skill for X" or "turn this into a skill". Not for merely using an existing skill.
- Skill JudgeUse when evaluating a pasted skill definition against a quality rubric -- knowledge delta, triggers, anti-patterns -- with scores and top fixes. Not for authoring new skills (use skill-creator).
- Skill Writing GuideUse when writing or editing the content of a skill file so it actually triggers and works: "improve this skill description", "why doesn't my skill fire", testing skills with subagents, skill structure, naming, and token efficiency. Not for scaffolding a brand-new skill file (use skill-creator).
More on Zeplik
Try Skill Self-Improvement on Zeplik
Every model, one chat. Bring the Skill Self-Improvement skill into your next conversation and let the assistant do the work.