Scaffold a new GitHub Copilot Agent Skill using the skill-creator skill workflow.
Scanned 9/4/2026
Install to Claude Code
npx -y skills add paulasilvatech/awesome-harness-primitives --skill commands --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Commands?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/paulasilvatech-commands)More formats (shields.io, HTML) on the badges page.
---
description: Scaffold a new GitHub Copilot Agent Skill using the skill-creator skill workflow.
argument-hint: skill name and what it should do
---
<!-- Generated from harness/github-copilot/prompts/new-skill.prompt.md by harness/claude-code/scripts/convert_from_copilot.py. Edit the source, not this file. -->
# /new-skill
## Objective
Scaffold a new GitHub Copilot Agent Skill under `harness/github-copilot/skills/` by routing the authoring workflow through the `skill-creator` skill and aligning the result with the current `skill.template.md` structure, validation expectations, progressive-disclosure model, and description quality bar.
## When to Invoke
Use this prompt when a user asks to create, scaffold, draft, or validate a new reusable GitHub Copilot Agent Skill for this repository.
## Preconditions
- The `skill-creator` primitive is available as a skill in VS Code.
- The requested skill name, purpose, and activation triggers are known or can be requested.
- Edits under `harness/github-copilot/skills/<name>/` are permitted.
- The author can add `references/`, `assets/`, or `scripts/` only when real content requires those bundled resources.
## Inputs the Team Must Provide
- Skill name — lowercase, hyphenated, and used as the folder name and frontmatter `name`.
- Purpose — what the skill does.
- Concrete triggers — keywords and phrases that should activate it.
- Whether the skill needs `references/` for on-demand Markdown, `assets/` for templates or static resources, or `scripts/` for runnable helpers.
- Ask the user for anything that is missing, and stop if the required authoring primitive is unavailable.
## What I Will Do
- Route the request to `skill-creator` (type: skill) before drafting or editing the skill.
- Create `harness/github-copilot/skills/<name>/SKILL.md` with valid frontmatter on line 1 and a folder name that equals `name`.
- Use the current mandatory skill spine: `## When to invoke`, at least one real domain section with a freely titled heading, `## Output template`, and `## Quality gate`.
- Include conditional sections such as `## Inputs`, `## Prerequisites and context`, `## Procedure`, `## Criteria`, `## Examples`, `## Limits`, `## Gotchas`, `## Troubleshooting`, `## Progressive disclosure and bundled resources`, `## Related primitives`, and `## References` only when their trigger is earned by real content.
- Keep `description` within 1024 characters, state what the skill does and when to use it, and enrich it with concrete trigger language.
## What I Will NOT Do
- Draft or edit the skill if `skill-creator` is unavailable in VS Code.
- Use the obsolete `## When to use` heading or a generic `## Workflow`/`## References` skeleton when the current template requires different sections.
- Add dangling references to `references/`, `assets/`, or `scripts/` resources that do not exist.
- Add sandbox home paths, `/mnt/...` paths, hard platform-product dependencies, unsupported metrics, or unverified claims.
- Narrate process steps in the final response when the requested output is a concise path and validation status.
## Output Format
Return only the concise creation result:
```markdown
### New Skill Result
### Skill Path
- `harness/github-copilot/skills/<name>/SKILL.md`
### Validation Status
- Frontmatter starts on line 1: `<passed|failed>`
- Folder name equals `name`: `<passed|failed>`
- Mandatory sections present in order: `## When to invoke`, `<domain section>`, `## Output template`, `## Quality gate`: `<passed|failed>`
- Conditional sections earned by real content: `<passed|failed>`
- Bundled resources under `references/`, `assets/`, or `scripts/` exist and are referenced only when needed: `<passed|failed|not applicable>`
- Script check: `<command and result, or not applicable>`
### Critical Findings or Blockers
- `<finding or none>`
```
## Definition of Done
- [ ] The request was routed to `skill-creator` before drafting or editing.
- [ ] `harness/github-copilot/skills/<name>/SKILL.md` exists and frontmatter parses from line 1.
- [ ] Folder name matches `name`, and `description` is rich with positive triggers and within 1024 characters.
- [ ] The delivered skill uses `## When to invoke`, at least one real domain section, `## Output template`, and `## Quality gate`.
- [ ] Every referenced file in `references/`, `assets/`, or `scripts/` exists, and any script is self-contained and runnable in this environment.
- [ ] The final response includes only the skill path, validation status, and critical findings or blockers.
## Prompt Body
Follow these steps in order. This prompt is a meta-primitive for authoring skills and must track the current skill template.
**Step 1 — Route through the authoring skill.** Load and use `skill-creator` (type: skill) before drafting or editing. If `skill-creator` is unavailable in VS Code, stop and report that the required authoring primitive is missing.
**Step 2 — Collect the required inputs.** Ask for missing skill name, purpose, activation triggers, and bundled resource needs. Validate that the name is lowercase and hyphenated. Treat it as the folder name and the frontmatter `name`.
**Step 3 — Create the skill location and frontmatter.** Create `harness/github-copilot/skills/<name>/SKILL.md` with frontmatter on line 1:
```markdown
---
name: <name>
description: "<what it does>. Use this skill when <triggers and keywords>."
---
```
Keep `description` within 1024 characters and use concrete positive trigger phrases.
**Step 4 — Use the current skill section map.** After the H1 and overview, include `## When to invoke`, at least one freely titled domain section that carries real rules, commands, patterns, examples, or decision criteria, `## Output template`, and `## Quality gate`. Add optional frontmatter such as `user-invocable`, `argument-hint`, or `allowed-tools` only for concrete need. Include `## Inputs` when `argument-hint` is set. Use `## Procedure` only when order is load-bearing. Use `## Criteria` for reviews or evaluations. Add examples, limits, gotchas, troubleshooting, progressive disclosure, related primitives, or references only when their trigger is real.
**Step 5 — Add bundled resources only when needed.** Add `references/`, `assets/`, or `scripts/` only if the skill needs on-demand Markdown, templates/static resources, or runnable helpers. Reference each resource from `SKILL.md` and avoid dangling links. Keep `SKILL.md` focused and push depth into `references/` for progressive disclosure. Make scripts self-contained, prefer the standard library, and ensure they run in this environment.
**Step 6 — Enforce authoring rules.** Keep the skill portable: no sandbox home paths, no `/mnt/...` paths, no hard platform-product dependency unless the skill truly requires it. Write in English. Use “GitHub Copilot” rather than bare “Copilot”. Do not use em dashes. Never fabricate metrics.
**Step 7 — Validate and report.** Confirm that the skill appears in the available-skills list when frontmatter parsing is available. Check that the folder name matches `name`, every referenced file exists, and any script runs. Return only the skill path, validation status, and critical findings or blockers.
## Invocation Example
```
/new-skill name=api-contract-review purpose="review API contracts" triggers="review OpenAPI, API compatibility, breaking changes"
```
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!