creatorskill

Purpose

This skill creates a lightweight, production-ready Pi resource (prompt template or Agent skill) from a short user brief. It favors concise, actionable files that follow the Agent Skills conventions and progressive disclosure: only the minimal metadata is kept in immediate context and larger references are stored separately.

When to use

• You want a new prompt template or skill scaffolded quickly from a short description.
• You want a small, standards-compliant SKILL.md written for immediate use by Pi or other Agent Skill runtimes.
• You prefer a single, reviewable change staged in chat (nothing is executed automatically).

Input contract

The skill expects a JSON payload (appended to the command) or equivalent user message with these fields:

• kind: "auto" | "prompt" | "skill" # preferred resource type or "auto" to infer
• scope: "global" | "project" # target location (global preferred)
• brief: string # one-paragraph description of what the resource should do
• name: string? # optional slug hint (will be normalized)

Behavior

1. Validate the brief. If essential details are missing, ask one concise clarification question.
2. If kind == "auto", infer whether a prompt template or skill fits the brief.
3. Determine a safe slug:
   • Prefer an explicit name if provided after normalization.
   • Otherwise derive a short, verb-led slug from the brief (lowercase, letters/numbers/hyphens, ≤64 chars).
4. Create the resource under the chosen scope using Pi conventions:
   • Prompt template: ~/.pi/agent/prompts/.md
   • Skill: ~/.pi/agent/skills//SKILL.md
5. SKILL.md structure for skills:
   • YAML frontmatter with name and description (these are the trigger fields)
   • Body: short Purpose, Setup (one-time steps), Workflow (ordered steps), Examples (if helpful)
   • Keep body concise; move large examples or references to references/ files.
6. For prompt templates:
   • Create a markdown file with frontmatter description and a short reusable prompt body and usage notes.
7. Never execute external commands or run the created scripts—always stage the operation and summarize the created files for user review.

Naming rules

• 1–64 chars, lowercase a-z, digits, hyphens only
• No leading/trailing hyphens, no consecutive hyphens
• Must match parent directory for SKILL.md (skill folder name)

Examples

Prepare a skill in chat (staged command):

/skill:creatorskill {"kind":"skill","scope":"global","brief":"Create a skill that reviews staged git changes for security issues and outputs a short remediation plan."}

Prepare a prompt template in chat (staged command):

/skill:creatorskill {"kind":"prompt","scope":"global","brief":"Template for reviewing staged git changes with a strict security checklist."}

Sample SKILL.md created for a security-review skill:

---
name: security-review
description: Review staged git changes for security issues and suggest concise remediation steps. Use when you want an automated checklist and file-level recommendations for changed files.
---

# Security Review

## Purpose
Quickly identify security-relevant changes in staged files and provide prioritized remediation suggestions.

## Setup
No one-time setup required. (If helper scripts are included, document how to run them.)

## Workflow
1. List staged files and their diffs.
2. For each file, check for secrets, insecure patterns, and risky configuration changes.
3. Produce a short summary and per-file remediation steps.

## Examples
- `./scripts/check_secrets.sh` (run locally after review)

Validation & Best Practices

• Keep SKILL.md focused: put large references under references/ and link from SKILL.md.
• Frontmatter MUST contain name and description (description should include trigger contexts).
• Prefer short, imperative language and examples instead of long prose.
• Ask at most one clarifying question if input is ambiguous.

Notes

• This skill only prepares and writes files under the chosen scope. It does not run or install anything.
• On successful creation, summarize exactly which files were written and their absolute paths so the user can review and run them manually.