# Author Skill

> Authors portable skills when asked to create or revise SKILL.md workflows and client distributions; preserves routing, provenance, and invocation controls rather than treating ordinary docs as skills.

- Skill: `jovanipink/author-skill` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx get-skill add jovanipink/author-skill`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jovanipink/author-skill/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: JovaniPink (https://skillmd.com/u/jovanipink)
- Updated: 2026-11-15
- Page: https://skillmd.com/skills/jovanipink/author-skill

---


# Portable Skill Authoring

Create one canonical workflow source and validate its client projections.

Read [client adapters](references/client-adapters.md) when invocation controls, packaging, or generated Codex, Claude, and Antigravity output are in scope.

## Workflow

1. Establish the capability gap, target requests, near misses, conflicting skills, supported clients, and authority boundary.
2. Confirm that the capability belongs in public core, a private overlay, or neither. Resolve license and source provenance before writing.
3. Choose a concise name and a discriminating description whose first clause carries the main trigger.
4. Keep the entrypoint focused on decisions the host would not make reliably without it. Move conditional detail into small, linked references.
5. Choose implicit or explicit invocation from workflow intent. Do not grant tools merely because the workflow might use them.
6. Add positive, near-miss, conflict or safety, output-quality, and installed-versus-baseline evaluations.
7. Generate native adapters and packages from canonical source. Validate schema, links, policy mapping, deterministic output, package layout, provenance, originality, and public boundaries.
8. Record observed behavior separately for every client surface. Do not infer parity from compatible files or another client result.

## Boundaries

- Prefer the host's native skill creator when the task is limited to that host, but preserve this catalog's portable acceptance contract.
- Do not copy a popular skill because its repository is public. Pin license and revision, review the full influence surface, and author independently when clean-room treatment is required.
- Do not bundle hooks, agents, MCP servers, dependencies, broad permissions, or executables without an explicit later release decision.

## Output

Report the capability, routing, invocation policy, canonical changes, adapters,
provenance, security, evaluations, validation, and observed client limits. These
are required meanings, not mandatory headings; match detail to the request.

Test concise and requested long-form outputs, competing instructions, partial
failure, stale handoffs, and interruptions. Keep shared meaning canonical and
native loading and invocation controls in generated adapters. A slash command
does not prove explicit-only enforcement; exclude those workflows from an
unverified surface.

Read [the original acceptance example](references/continuity-example.md) when checking this behavior.

## Example

Synthetic example; not an observed client result.

Input: A client has a manual skill command but no verified implicit-selection control.

Output: Keep explicit-only workflows unavailable on that surface until the control is verified. File compatibility does not establish behavioral parity.

