Package Skill

Create and manage skill packages — bundles of related skills merged into a single top-level skill with internal dispatch. A package skill's own description d...

Security Scan

The skill's code and instructions match its stated purpose (packaging other skills) and do not request credentials or network access, but it performs destructive filesystem operations (moving/deleting other skill directories) so you should review and back up your skills before using it.

Purpose & Capability

Name/description, SKILL.md instructions, and the included scripts all implement a skill-packaging tool: copying sub-skills into a package, generating pack.md, producing a package SKILL.md, and removing/restoring top-level skill directories. The requested capabilities (filesystem operations under the skills/ tree) are coherent with the stated purpose.

Instruction Scope

SKILL.md instructs the agent/operator to read pack.md and individual SKILL.md files and to run the provided create/update scripts. The scripts explicitly remove the original top-level skill directories during 'create' and 'add' operations (shutil.rmtree/shutil.move). Reading and modifying SKILL.md and pack.md is expected, but the destructive removal of original directories is a significant action that the user should be aware of.

Install Mechanism

No install spec or third-party downloads; the skill is instruction-only with included Python scripts. Nothing is fetched from external URLs and no packages are installed automatically, which minimizes supply-chain risk.

Credentials

The skill requests no environment variables or credentials. It does, however, require read/write access to the local skills/ directory (calculated relative to the script location). That filesystem access is necessary for packaging work but is a privilege to consider.

Persistence & Privilege

Although always:false, the scripts change other skills on disk (copy, remove, and restore top-level skill directories). Modifying or deleting other skill directories is within the tool's purpose but is a high-impact operation — it can permanently remove top-level skills if used incorrectly, so exercising caution and backups is important.

Assessment

This skill is internally coherent and appears to do what it says, but it performs high-impact filesystem changes (copies, moves, and deletes of skill directories). Before installing or running it: 1) Review the two Python scripts yourself (they are included) to confirm behavior; 2) Back up your skills/ directory so you can recover accidentally removed skills; 3) Test on a copy of your skills directory or in an isolated environment; 4) Prefer running the CLI manually (python .../create.py ...) rather than allowing automated/unsupervised agent invocation; 5) Ensure you use only safe package and skill names (the scripts validate names, but avoid running with names you don't trust). No network or credentials are required by the tool.

Skill Packager

Group related skills into a single package with internal dispatch. Reduces top-level skill count and trigger noise while preserving all capabilities.

When to Use

• Multiple skills share similar trigger patterns (e.g., docx/pptx/xlsx/pdf all trigger on file extensions)
• Skill count exceeds 25 and trigger accuracy degrades
• User asks to consolidate or merge skills

When NOT to Use

• Skills have fundamentally different trigger domains (e.g., "weather" and "github" — keep separate)
• Only 1-2 skills are involved — not worth packaging overhead

Directory Structure

skills/
  package-skill/           ← This tool
    SKILL.md
    scripts/
      create.py            ← Create packages, add/remove sub-skills
      update.py            ← Scan sub-skills, update pack.md
  <package-name>/          ← Generated package (independent skill)
    SKILL.md               ← Trigger + sub-skill list + dispatch logic
    pack.md                ← Registry: auto-generated sub-skill inventory
    sub/
      <skill-a>/SKILL.md
      <skill-b>/SKILL.md

Workflow

1. Create a Package

python skills/package-skill/scripts/create.py create <package-name> -d "trigger description" <skill1> [skill2 ...]

This will:

• Create <package-name>/ directory with sub/ and pack.md
• Copy each named skill into sub/
• Generate a SKILL.md with the provided trigger description and dispatch instructions
• Generate a pack.md with all sub-skill descriptions for matching
• Remove the original top-level skill directories

2. Update pack.md

python skills/package-skill/scripts/update.py <package-name>

Re-reads all sub/*/SKILL.md frontmatter and updates pack.md. Use after adding or removing sub-skills manually, or anytime you want to refresh the registry.

3. Add a Sub-Skill to Existing Package

python skills/package-skill/scripts/create.py add <package-name> <skill-name>

Copies a new skill into the package's sub/ and updates pack.md.

4. Remove a Sub-Skill

python skills/package-skill/scripts/create.py remove <package-name> <skill-name>

Removes a sub-skill from the package. If the sub-skill deserves to be top-level again, this command moves it back to skills/ root.

Package SKILL.md Template

The generated SKILL.md follows this structure:

---
name: <package-name>
description: "<user-provided trigger description>"
---

# <Package Title> Package

This is a skill package bundling N sub-skills: skill-a, skill-b, ...

## How to Use

1. Read [pack.md](pack.md) to determine which sub-skill matches the current task
2. Read the matching sub-skill: `read_file("skills/<package-name>/sub/<sub-skill>/SKILL.md")`
3. Follow that sub-skill's instructions

Description is concise and provided by the user at creation time. No sub-skill descriptions here — they live in pack.md.

pack.md Format

# <package-name> Registry

- **pdf** — Use this skill whenever the user wants to do anything with PDF files...
- **docx** — Use this skill whenever the user wants to create, read, edit Word documents...

Each entry is the name and description from the sub-skill's frontmatter. Auto-generated by create.py scan.

Guidelines for Writing Package Descriptions

A good package trigger description should:

1. Cover all trigger keywords (file extensions, actions) from sub-skills
2. Stay under 200 words — the description is loaded in every session
3. Guide the agent to read pack.md for sub-skill matching

Design Principles

• Sub-skills stay intact — no content is stripped, only relocated to sub/
• Package is a normal skill — the agent treats it like any other skill
• Dispatch is explicit — the agent reads the sub-skill SKILL.md on demand, not pre-loaded
• Reversible — remove can promote a sub-skill back to top-level