# Learn

> >

- **Type:** Skill
- **Install:** `agentstack add skill-bioinfo-claudelicious-learn`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [BioInfo](https://agentstack.voostack.com/s/bioinfo)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [BioInfo](https://github.com/BioInfo)
- **Source:** https://github.com/BioInfo/claudelicious/tree/main/skills/learn

## Install

```sh
agentstack add skill-bioinfo-claudelicious-learn
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# learn

Turn an external artifact into an adoption decision. The output is not a summary
and not a grade. It is a short list of the things worth lifting into your own
stack, each with a decision attached. The judgment that makes this skill useful is
that it reads the artifact **against what you already run**, not in a vacuum, so
it never recommends something you already do under a different name.

## What this is NOT

- Not a review or a decode. Those grade an artifact on its own merits. This grades
  the artifact against your systems and outputs a lift plan.
- Not a full re-implementation. It decides *whether* and *how* to adopt; the
  building happens after the decide-gate says so.
- Not a summarizer. If it hands you a description of the artifact and stops, it
  did the wrong job. Every run ends in decisions.

## Procedure

1. **Resolve the artifact.** Fetch and read the link closely enough to know what
   it actually claims, not just its headline. A repo means the README plus the
   parts that carry the real idea; a paper means the method, not the abstract; a
   thread means the argument, not the first post.
2. **Map against what you already have.** Search your own stack for overlap: your
   notes, the skills folder, your rules, your memory. The goal of this step is to
   separate a genuinely new idea from one you already run. Skip it and you will
   recommend things you already do.
3. **Extract the deltas.** Keep only what the artifact does that you do not.
   Discard everything you already cover. A short list of real deltas beats a long
   list padded with things you already have.
4. **Red-team each delta against your real setup.** For every surviving delta,
   ask not "is this good in the abstract" but "does it survive contact with how I
   actually work, and what breaks if I adopt it." A delta that fails this step is
   a skip, and saying so is a result.
5. **Decide, per delta.** Attach one of four calls to each surviving delta:
   **borrow** the idea and adapt it, **build** your own version, **install** the
   thing directly, or **skip** it with the reason.

## The decide-gate

Close with a single gate rather than an open recommendation. For the deltas worth
acting on, offer three exits: **plan it further** (the change is real but needs a
design pass), **borrow it now** (small enough to lift in place), or **implement
it** (start building against the decision). The user picks the exit. The skill
does not silently start building.

## Notes

- Weight what red-teaming survives, not what sounds clever. A delta that reads
  well but breaks against your setup is the weakest kind of recommendation.
- Be willing to conclude "nothing to lift here." An artifact you already cover is
  a valid, useful answer, and it is faster than a padded lift plan.
- Keep the deltas honest to the source. If a capability is not actually in the
  artifact, it does not go in the plan.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [BioInfo](https://github.com/BioInfo)
- **Source:** [BioInfo/claudelicious](https://github.com/BioInfo/claudelicious)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-bioinfo-claudelicious-learn
- Seller: https://agentstack.voostack.com/s/bioinfo
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.