skill-creator
Generate a new skill from a plain-language description — decides invocation control, arguments, and context cost, then scaffolds, validates, and tests it
conorbronsdon · Static checks on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add conorbronsdon/agent-context-os --skill skill-creatorAbout this skill
name: skill-creator description: Generate a new skill from a plain-language description — decides invocation control, arguments, and context cost, then scaffolds, validates, and tests it x-source: "maintainer-core/skills/skill-creator/SKILL.md" x-source-version: "d66f695"
skill-builder — Skills That Don't Rot
Takes a plain-language description of a task and produces a ready-to-ship agent skill. The design goal is skills that are cheap in ambient context and current with the skill spec — most skill generators fail on exactly those two axes: they don't know about arguments/argument-hint, never suggest disable-model-invocation, and write bloated descriptions that tax every session.
Modes
Parse $ARGUMENTS for a mode; default to new when the input is a description.
new <description>— scaffold a skill from scratch (the flow below).review <path>— audit an existing SKILL.md against steps 1–3's decision points and run the validator (step 4). Report findings; don't rewrite unless asked.migrate <path>— convert a legacy.claude/commands/*.mdfile into a skill directory: carry the frontmatter over, apply the review checklist, and place supporting content in the directory.
Before You Start
Read references/claude-code-frontmatter.md — a pinned, dated snapshot of the skill frontmatter spec, argument substitutions, and context-loading behavior.
Staleness policy: if the snapshot date is more than 3 months old, warn the user and continue with the pinned snapshot. Do not re-fetch or modify the reference mid-generation — updating the spec is a maintenance operation (agent-skill-builder's spec-drift CI watches for it), not a side effect of building a skill. Only update the snapshot when the user explicitly asks.
Also read the host repo's CLAUDE.md and existing skills — avoid duplicating a skill that already exists, and match local conventions.
Instructions
1. Clarify the skill
Ask (or infer from context):
- Target runtime: Claude Code (default — full frontmatter feature set) or the portable Agent Skills core (
name+descriptiononly; keep Claude-specific fields out or document them as optional extensions). - Name: lowercase, hyphenated. In Claude Code the directory name under
.claude/skills/is the command you type; frontmatternameis only a display label. - Who invokes it — the most important design decision:
- User-only (side effects, or timing the user controls: publish, send, deploy, log) →
disable-model-invocation: true. Bonus: the description is then not loaded into every session — zero ambient context cost. - Claude-only (background knowledge, not an action) →
user-invocable: false. - Both (default) → the description must carry the trigger keywords, and it becomes a permanent per-session context cost. Make it earn that.
- User-only (side effects, or timing the user controls: publish, send, deploy, log) →
- Arguments: does the user pass input on the command line? If yes: set
argument-hint, and consume$ARGUMENTS(or$0/$1, or namedarguments:) in the body. Don't make the model guess what the trailing text means. - Input/Output: what does it read (files, MCP data, live command output), what does it produce?
- Tools:
allowed-toolsis a pre-approval grant (no permission prompts while active), not a restriction — scope entries tightly, e.g.Bash(git add *)notBash. Usedisallowed-toolsto actually remove tools. - Where it runs: inline (default — shares conversation context) or
context: fork+agent:for isolated, self-contained tasks. Fork only works when the body is an explicit task, not reference guidelines. - Data at load time: use dynamic context injection (
!`command`at line start) to inline live data (a diff, a status file) before the model reads the skill. - Only set
model/effortoverrides with a specific reason. Usepaths:for skills scoped to certain files.
2. Write the description — the context-budget step
- 1–2 sentences, key use case first, third person, containing the words the user would actually say. Target ~250 chars; the listing truncates at 1,536.
- Detail, steps, and edge cases go in the body — the body loads only on invocation, so it's nearly free until used. Never pad the description to be "thorough"; that's the anti-pattern that bloats every session.
- If
disable-model-invocation: true, the model never sees the description — write it for the human scanning the/menu, and keep it short.
3. Draft the SKILL.md
Frontmatter template — include only the fields this skill needs (all are optional; see the reference for the full table):
---
name: [skill-name]
description: [1-2 sentences, triggers first]
argument-hint: "[expected args, e.g. [issue-number] [format]]" # only if it takes arguments
disable-model-invocation: true # only for user-triggered workflows
allowed-tools: [tightly-scoped grants] # only if prompts would be annoying
---
Body structure: short purpose line, "When to Use" bullets, numbered step-by-step instructions, explicit output format.
Quality checklist:
- Steps, not paragraph prose; each step names the tool or command
- Output format defined, not left ambiguous
- Instructions written as standing rules — skill content stays in context for the rest of the session, so "do X once" phrasing goes stale; "always X when Y" doesn't
- Body under 500 lines; long reference material moved to supporting files in the skill directory, linked from SKILL.md
- No hardcoded IDs, paths, or credentials — use config files, env vars, or
${CLAUDE_SKILL_DIR}for bundled scripts - Confirmation gate on anything that changes external state (git push, API calls, file deletes)
4. Validate
If agent-skill-builder is ava
Read the complete source ↗