George O'NairField Lab

Shop / Developer Tooling / Agentic Distiller

AGENTIC-DISTILLER · PREVIEW
Agentic Distiller

Agentic Distiller is a local Python package for reviewing a repository and starting an Agent Skill. Its deterministic CLI records license and repository signals, creates a SKILL.md skeleton with placeholder files, and checks the package structure. You or your coding agent must implement the actual workflow and test it on your own repository before use.

Target user

  • Builders who can inspect a local repository and finish a skill skeleton
  • Developers who want a repeatable starting point and will test the resulting workflow

Non-goals

  • Anyone expecting automatic source-to-working-tool conversion
  • Teams requiring a secret scanner, security certification, or verified agent-runtime integration
  • Hosted SaaS workflows requiring cloud API hosting, recurring subscriptions, or managed compute
  • Users expecting third-party external code auditing or manual consulting from the publisher
What the package does

Audit, scaffold, validate

Run the local Python CLI on a repository you are allowed to inspect. It records license and repository signals, creates a skill skeleton, and checks basic package structure. The generated scripts are placeholders until you implement and test a real workflow.

Step 01

Review the local repository

Inspect license class, documentation, tests, examples, and file inventory. Treat the verdict as a signal for your own review.

Output: preflight verdict and repository inventory
Step 02

Create a skill skeleton

Generate SKILL.md plus script, template, reference, and example stubs. Source code is not converted into working tools.

Output: a directory with TODO placeholders
Step 03

Validate the package

Check frontmatter, Python compilation, and restricted-license terms. A passing check does not test the intended business workflow.

Output: CLI PASS or review findings
Reproducible fixture

The included sample run

From the downloaded package root, run the included sample repository. The fixture has produced status PASS locally; this verifies scaffolding and packaging checks, not a finished Agent Skill.

TERMINAL — SAMPLE REPOSITORYLOCAL PYTHON CLI
python3 skills/github-repo-skill-distiller/scripts/distill_local_repo.py   --repo fixtures/sample-todo-cli   --name todo-cli-maintainer   --output /tmp/sku03-out

Expected: JSON status "PASS"; generated SKILL.md and TODO script stubs.

To make the output useful, choose one job, replace the TODO files with real commands and instructions, then test it in the agent you use. Agent installation paths are documented, but automatic discovery has not been independently verified across hosts.

Before you buy

Limits to know

Local path required

The CLI does not clone a GitHub URL. Give it a repository already present on your machine.

Manual implementation required

It does not extract AST tool contracts, write working business logic, or create a runtime test harness.

Review private material

It does not certify security or automatically scrub secrets. Review the source and generated files before sharing them.

Check agent integration

Skill directory paths are documented. Claude Code, Codex, Cursor, and OpenCode discovery remain unverified in this package.

Technical FAQ

Frequently asked questions

Does the CLI extract working tools from source code?

No. It inventories and audits a local repository, then creates a skill skeleton. The script and template files are placeholders. A human or coding agent must choose a workflow, implement it, and test it.

What is the difference between Agentic Distiller and tools like skills-ref or npx skills?

Agentic Distiller helps start a skill from a local repository by recording preflight signals and creating a skeleton. It does not install skills for you or generate a working tool from source.

Does the CLI remove secrets or guarantee safe agent execution?

No. It does not copy source code into the generated skill, but it is not a credential scanner or security guarantee. Review the repository, generated files, permissions, and any instructions before running or distributing them.

Can I use it with a private or multi-language repository?

Point the CLI at a local directory you are authorized to inspect. It records repository signals rather than parsing each language into working tools. Review licensing and private files yourself before sharing any output.

What artifacts are included in the $29 one-time download?

The package includes the github-repo-skill-distiller CLI, book-skill-distiller templates, a sample-todo-cli fixture, installation instructions, and documentation. The generated skill needs additional implementation.

Technical specifications

Input
A local repository you are authorized to inspect
CLI
Python 3; fixture smoke-tested on macOS arm64
Output
SKILL.md skeleton and placeholder support files
Validation
Frontmatter, Python compile, restricted-license terms
Agent installs
Paths documented; runtime discovery not independently verified
Required next step
Implement and test the real workflow yourself or with an agent

What the product listing includes

Local Python CLI for license and repository preflight, scaffolding, and package validation
Skill skeleton with SKILL.md and script, template, reference, and example stubs
License verdicts: DISTILL, INTERNAL_ONLY, NEEDS_REVIEW, or BLOCKED
Checks for frontmatter, Python compilation, and restricted-license terms
book-skill-distiller templates for agent-guided manuscript work
sample-todo-cli fixture and expected output for a local smoke test