Askit Build Skill

askit-build-skill

Purpose

Author conformant skills. Two modes: create scaffolds a new skill that passes Bronze on first run; improve raises an existing skill toward the quality bar. Authoring depth is in [references/authoring-guide.md](references/authoring-guide.md).

When to use

When the user asks to create, scaffold, write, or improve a skill.

create mode

1. Brief interview: ask for the skill name (kebab-case), what it does, when to use it, and a few trigger keywords. Skip the interview if these inputs are already provided in context.
2. Delegate the drafting to the askit-skill-author subagent (the declared chain) when running in a harness that can dispatch it; otherwise perform the steps below inline. Either way the procedure is the same: create skills// and copy templates/SKILL.md into skills//SKILL.md.
3. Fill the frontmatter: name equal to the directory, and a description that states what AND when with concrete keywords (see [references/authoring-guide.md](references/authoring-guide.md) for the bar).
4. Scaffold references/ if the skill needs depth. Samples in examples/ are optional at Bronze - add them only if useful. Do not assume the surrounding plugin anatomy exists - this skill works a la carte.
5. Emit native manifests for the declared targets: set agent-targets in the plugin's library.json (for example ["claude", "codex"]), then run node scripts/generators/gen-manifest.mjs --write --target=all. To emit for one agent only, pass --target=claude or --target=codex. The plugin's library.json agent-targets (not a CLI flag) declares which targets it requires; gen-manifest writes .claude-plugin/plugin.json and/or .codex-plugin/plugin.json from library.json.
6. Assess the new skill with node scripts/evaluate.mjs skills/ --json (this is what the askit-evaluate skill runs), report the result, and iterate until it passes with 0 errors.

improve mode

1. Run node scripts/evaluate.mjs --json and read the report. As in create mode, the rewrite work itself may be delegated to askit-skill-author or done inline; the steps below are the procedure either way.
2. For each finding: a samples warn -> add representative examples; a low description score (U5) -> rewrite the description to clear the bar; an over-budget warning (U7) -> move depth into references/. For any other finding, read its message and apply the fix it states.
3. Re-run evaluate to confirm the findings are resolved.

Scope

Emits for Claude and Codex: library.json.agent-targets declares which targets the plugin requires, and gen-manifest.mjs --target=all|claude|codex generates the matching native manifests. This skill delegates authoring to the askit-skill-author subagent (permitted in agents/_chain-permitted.yaml), which in turn delegates assessment to askit-evaluator.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: product-on-purpose
• Source: product-on-purpose/agent-skills-toolkit
• License: Apache-2.0
• Homepage: https://product-on-purpose.github.io/agent-skills-toolkit

Install and usage instructions live in the source repository linked above.