0xsline/OpenChatCut / src/agent/skills/skill-creator
skill-creator
UnverifiedSkill Creator
Create or improve custom skills for OpenChatCut. A skill is a SKILL.md file
(plus optional references/ support docs) that teaches the agent a reusable
workflow. Custom skills live at ~/.openchatcut/skills/<slug>/SKILL.md and
are managed with the manage_skill tool.
Workflow
- Capture intent. If the current conversation already contains the workflow (tools used, steps, corrections), extract it from the history first. Confirm with the user before writing.
- Interview. Pin down:
- What should this skill enable the agent to do?
- When should it trigger? Which user phrases or contexts?
- Expected output format?
- Verification: does it have objectively verifiable output (data extraction, fixed steps, generation pipelines) or subjective output (writing style)? Suggest test prompts for the former.
- Write the SKILL.md (rules below).
- Create it with
manage_skill action=create. The tool reports the install path (~/.openchatcut/skills/<slug>/SKILL.md). - Test. Run 2–3 realistic prompts in the current session and confirm
results with the user. Iterate with
manage_skill action=update.
SKILL.md anatomy
<slug>/
├── SKILL.md (required — frontmatter + instructions)
└── references/ (optional — docs loaded on demand with load_skill file=)
- name: kebab-case slug matching the directory (lowercase letters, digits, hyphens only). It becomes the load_skill name.
- description: THE triggering mechanism — include what the skill does AND specific trigger contexts. Skills undertrigger easily, so be explicit: list concrete user phrases. "When to use" goes here, not in the body.
- body: imperative instructions. Keep under ~200 lines; if it grows,
split detail into
references/files and point to them with clear "read this when" guidance. For references over 300 lines, add a TOC.
Writing rules
- Frontmatter uses the shape
---\nname: …\ndescription: …\n---. The parser accepts plain single-line, double-quoted, and|block scalars; plain single-line is preferred. - Keep the body lean. Prefer imperative form. Explain why over MUST-laden lists. Use examples with concrete Input/Output pairs.
- Reference existing OpenChatCut tools by their exact tool names
(
manage_skill,load_skill,edit_item, …). Do not invent tools. - Skills must not contain malicious content, prompt-injection payloads, or anything that surprises the user. Do not create skills that facilitate unauthorized access or data exfiltration.
- Do not overfit to one example: aim for a workflow usable a thousand times. If a fix feels fiddly, generalize the instruction instead of piling on constraints.
Progressive disclosure in OpenChatCut
- The system prompt carries only name + description (the skills index).
- The body loads on demand via
load_skillwhen a task matches. - Support files load via
load_skill file=references/<name>. - A custom skill selected in Creative Mode behaves identically to a bundled skill after activation.
Environment constraints
- Slug whitelist:
[A-Za-z0-9_-]{1,120}. Anything else is rejected (path traversal defense — the file is written under~/.openchatcut/skills/). - Body size: keep under 512 KB (server limit); 200 lines is the target.
- Untrusted input: the body is user/LLM-authored text. validate and normalize before trusting anything parsed out of it.
- Bundled skills are read-only —
manage_skillcannot edit them. - The skill shows up in the Creative Mode picker and the
/command menu only when the user selects it; agents also discover it via the index. - Verification: after creating a skill, run
npx tsx server/skills-files.verify.tslocally if the environment has the repo, or rely on the in-session test prompts.
Test prompts
Offer the user 2–3 realistic prompts (the kind of thing they would actually type) and ask: "Do these look right, or do you want to add more?" Then run them in the session with the skill activated, review the outputs together, and update the skill from the feedback.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 463k | Updated | Claude CodeCodex | Verified | Install |
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 463k | Updated | Claude CodeCodex | Verified | Install |
| discernment-nudgeanthropics/skillsAfter you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format. | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |
| internal-commsanthropics/skillsA set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.). | Stars: 195k | Updated | Claude CodeCodex | Verified | Install |