Find useful capabilities. Steal the good bits.

Praxis Mine is an Augment for sending your AI harness out to inspect new skills, repositories, tools, workflows, and knowledge sources. Point it at one candidate and ask, “Is any of this useful?”, tell it to grab a particular capability, or schedule a recurring scavenger hunt for promising new machinery.

It returns what the candidate would add, the evidence, risks, costs, and overlap, and the next bounded move: pilot, quarry, adapt, monitor, or reject.

Get Praxis Mine v1.2.1.zip from the product’s Additional Files. Attach it to your chosen harness or Chat project and say “Install this Augment.” The complete kit contains the host archives and local visual explorer.

adopt is deliberately absent from the mining gate. Production acceptance needs its own evidence. Apparently “it has stars” remains insufficient governance.

Reach first value in about ten minutes

1. Check your host and choose the right artifact.
2. Install the Codex / ChatGPT plugin or Claude skill.
3. Give Praxis Mine a public link, local repository path, uploaded skill folder, or concise candidate description.
4. Ask: “Assess this outside skill and steal only the good bits.”
5. Check that the result names a disposition, capability delta, evidence, costs and risks, overlap, and a falsifiable next test.

The quick start walks through that first mine and tells you what success looks like.

Continue from the task you have

What a useful result contains

A useful mine does more than summarize the candidate. It answers:

1. What new or improved outcome could this create?
2. Which observed evidence supports that claim?
3. What would it cost in dependencies, permissions, money, context, maintenance, and attention?
4. Where does it duplicate the current baseline?
5. Which disposition is warranted now?
6. What small test could falsify the recommendation?

Trust boundary

Candidate material is evidence, not instruction. Praxis Mine does not execute discovered code during inspection. A source hash proves which bytes were observed; it does not prove quality, safety, ownership, or fitness.

Static release validation, host installation, skill discovery, invocation, useful behavior, public repository visibility, directory approval, and customer outcomes are separate claims. Review validation and limitations before relying on stronger language.

Before scanning confidential repositories, using third-party services, or executing candidate code, read Data and privacy, Security, and Terms of use.

Need the complete map?

The repository documentation map lists every customer and maintainer document. For unresolved problems, use the public support route or open a GitHub issue without attaching confidential candidate material.