Skillpack Creator

Overview

Turn a successful task into a reusable SkillPack. Extract the stable workflow, decide what belongs in a local skill versus pack-level prompts, generate the pack structure, and package it only after the workflow is explicit and repeatable.

Workflow

1. Normalize the source task

Start by reducing the finished task into a clean execution spec:

Capture the user goal and the concrete deliverable.
Capture the final successful workflow, not the full exploratory transcript.
List required skills, tools, files, secrets, and environment assumptions.
Separate deterministic steps from heuristic steps.
Remove dead ends, retries, and one-off debugging noise.

If the prior work is still ambiguous, ask for the missing stable facts or infer only the pieces that are low risk.

2. Decide what the pack should contain

Use this split:

Put reusable procedural knowledge in a local skill under the target pack's skills/.
Put repeated shell or file-generation logic into scripts inside that local skill when reliability matters.
Put detailed schemas, API notes, or conventions into reference files.
Put 1 to 3 pack-level prompts in skillpack.json as the pack's user-facing entry points.

Do not treat prompts as a strict workflow engine. In this codebase they are starter inputs for the UI, not a DAG or state machine. Read references/skillpack-format.md when you need the exact pack semantics.

3. Create the pack specification

Before writing files, define:

Pack name
Pack description
Preset prompts
Skill list with name, source, and description
Which skill is the new local orchestrator skill
Expected output files and success criteria

Prefer one local orchestrator skill plus a small number of external skills. Keep the pack narrow and job-focused.

4. Create the local orchestrator skill

In the target pack:

Create skills/<skill-name>/SKILL.md.
Write frontmatter that clearly describes what the local skill does and when it should trigger.
Move the stable workflow into imperative instructions.
Add scripts/ only for fragile or repeated operations.
Add references/ only for detailed information that should not bloat SKILL.md.

If the workflow is mostly instructions, keep the local skill simple. If reproducibility depends on exact file generation, add scripts.

5. Materialize the pack

Use scripts/scaffold_skillpack.py in this skill when you already know the pack spec. It will:

validate a JSON manifest shaped like skillpack.json
write skillpack.json
create skills/
copy start.sh and start.bat from this repository's templates/
optionally run npx -y @cremini/skillpack zip

Typical usage:

python3 skills/skillpack-creator/scripts/scaffold_skillpack.py \
  --manifest /tmp/skillpack.json \
  --output /absolute/path/to/output-pack

With zip generation:

python3 skills/skillpack-creator/scripts/scaffold_skillpack.py \
  --manifest /tmp/skillpack.json \
  --output /absolute/path/to/output-pack \
  --zip

6. Validate the result

Before handing the pack back:

Confirm the manifest matches the intended pack scope.
Confirm every declared skill has a valid name, source, and description.
Confirm local skills are present under the target pack's skills/.
Confirm the starter prompts are concrete enough to reproduce the workflow.
Zip only after the pack can already run as a directory.

Decision Rules

If the reusable value is mostly workflow knowledge, create a local skill and keep scripts minimal.
If the task depends on exact file emission or repetitive shell steps, script those parts.
If the task is still too broad, split it into a narrower pack instead of writing a vague mega-skill.
If key success conditions depend on hidden human judgment, state that the pack is a best-effort assistant workflow, not a deterministic pipeline.

Output Standard

When you use this skill, produce:

A short summary of the stabilized workflow.
The target pack structure and skill inventory.
The created or updated local skill files.
The generated skillpack.json.
Whether the pack was zipped and where the zip lives.