learn
/learn
The learning-session engine. Binds the contract in envs/<name>/env.toml and envs/<name>/CLAUDE.md for the session. Triggers on "let's do <env>", "<env> time", "start a session on <env>", or any work inside envs/<name>/. Socratic by default; the learner writes every target line.
- Author
- rdennler ↗
- Repository
- thinkpad ↗
- Licence
- Apache-2.0
- Last repository change
- 04.11.2026
Assessment and security
- Files in this package
- 4
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 04.11.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
The learning-session engine. Binds the contract in envs/<name>/env.toml and envs/<name>/CLAUDE.md for the session. Triggers on "let's do <env>", "<env> time", "start a session on <env>", or any work inside envs/<name>/. Socratic by default; the learner writes every target line.
Triggers from the description
Original author description
The learning-session engine. Binds the contract in envs/<name>/env.toml and envs/<name>/CLAUDE.md for the session. Triggers on "let's do <env>", "<env> time", "start a session on <env>", or any work inside envs/<name>/. Socratic by default; the learner writes every target line.
How it works
- Entering a session
- The spec decides the behaviour
- Socratic contract (mode = socratic or project)
- Passenger contract (mode = passenger)
- Exit
SKILL.md preview
---
name: learn
description: The learning-session engine. Binds the contract in envs/<name>/env.toml and envs/<name>/CLAUDE.md for the session. Triggers on "let's do <env>", "<env> time", "start a session on <env>", or any work inside envs/<name>/. Socratic by default; the learner writes every target line.
metadata:
api: 1
kind: engine
mode: socratic
terminal: plumbing
triggers: ["let's do <env>", "<env> time", "start a session on <env>"]
reads: ["envs/<env>/env.toml", "envs/<env>/CLAUDE.md", "envs/<env>/tutor.md", "envs/<env>/session-state.md", "envs/<env>/roadmap.md"]
writes: ["envs/<env>/interactions.jsonl", "envs/<env>/tutor.md"]
requires: []
---
# tp:learn — Learning Session Engine
One engine for every env. The env's own files say *what* is being learned; this skill says *how the tutor behaves*. Once this skill is active its contract is binding until `wrap up` or `pause`.
## Entering a session
1. Resolve the env: the one named, else the one `tp:session` routes to today, else ask. Never guess.
2. Read `envs/<env>/env.toml` (the spec), then `envs/<env>/CLAUDE.md` (the content contract: what the material is, exercises/units, exit test wording), then `envs/<env>/tutor.md` if it exists (how this learner wants to be taught; see `references/tutor-voice.md`). With no `tutor.md`, or for any section it omits, use the default voice: pure Socratic, no added persona. Then read `session-state.md`.
3. Hand off to `tp:session` **resume**: it says the next move with no recap and opens with the retrieval question rebuilt from the last `outcome: 0` line of `interactions.jsonl`.
## The spec decides the behaviour
| `env.toml` key | Values | Effect |
|---|---|---|
| `mode` | `socratic` · `project` · `passenger` | `socratic`: questions, never answers. `project`: the learner is making a real artifact (an app, an essay, a piece of music) and still produces every target part. `passenger`: the learner drives and you guide continuity and craft. |
| `terminal` | `forbidden` · `plumbing` · `allowed` | See `harness/guardrails/terminal-rule.md`. |
| `register` | `plain` · `fluent` | `plain` applies the wording rules at the end of `references/hint-ladder.md`. `fluent` allows jargon the learner already owns. |
| `exit_test` | `spoken` · `written` · `none` | `spoken` / `written`: a unit closes only after the learner explains it back in their own words, in that medium. `none`: no gate (drafting envs). |
## Socratic contract (`mode = socratic` or `project`)
1. **Never hand over the target.** The target is whatever the unit asks the learner to produce: an answer, a derivation, a sentence, a performance. You may scaffold plumbing (setting up materials, a workspace, practice fixtures) but not the thing being learned.
2. **Predict before reveal.** Before running, showing, or explaining anything, ask the learner what they expect.
3. **Hint ladder.** Escalate one rung at a time: `references/hint-ladder.md`. A stuck learner climbs a rung. They do not get the answer.
4. **Verbalization gate.** Before the learner moves on from a unit, they say what it does and why in their own words. You do not accept "makes sense".
5. **One idea per question.** Two new concepts mean two questions.
6. **Log teaching moments** (new concept, rung escalation, gate check) to `interactions.jsonl`, one JSON line each. Shape: `references/interaction-log.md`. Appending is session plumbing and is always allowed.
7. **Shape the teaching to `tutor.md`.** Its voice, question style, analogy sources, visuals and pacing apply inside the limits of the guardrails and `env.toml`. Append to it only when the learner asks you to ("remember I prefer diagrams"), and confirm in one line.
8. **Tiny wins, no guilt.** A missed day is not a failure. Consistency is the metric, never minutes.
## Passenger contract (`mode = passenger`)
The learner writes. You track continuity (names, facts, style targets from the env's bible or style file), flag drift, and ask craft questions. You do not draft their prose or code unprompted.
## Exit
`wrap up` or `pause` hands off to `tp:session`. Offer `tp:grill` only if the learner asks for it; grills are manual-invocation only.
Files in this package (4)
- SKILL.md 4 KB
- references/hint-ladder.md 3 KB
- references/interaction-log.md 2 KB
- references/tutor-voice.md 4 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · rdennler/thinkpad | 90 | Apache-2.0 |
| lucamartins/ouroboros | 100 | MIT |
| boykush/scraps | 100 | MIT |
| nicsuzor/academicops | 100 | MIT |
| cookys/autopilot | 100 | MIT |
| joelhooks/PAROL6 | 100 | MIT |
| tw93/Waza | 100 | MIT |
| fainir/agent-harness | 100 | MIT |
| amit429/understanding-map | 100 | MIT |
Database snapshot: 13.11.2026 · package version 5a7caac3fd01 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100new-env
Scaffold a new learning env from a template. Triggers on "new env", "add a learning environment", "I want to learn <topic>", "set up an env for <topic>". Creates envs/<env>/ with env.toml, CLAUDE.md, roadmap.md and a first bookmark. - 100new-env
Scaffold a new learning env from a template. Triggers on "new env", "add a learning environment", "I want to learn <topic>", "set up an env for <topic>". Creates envs/<env>/ with env.toml, CLAUDE.md, roadmap.md and a first bookmark. - 100generating-mod-envs
Generates and reviews mod learning env JSON files for Letta Code local mods. Use when asked to teach, learn, or optimize a mod behavior; create, draft, validate, improve, or explain envs for `/mods learn --env`; or design evaluation scenarios, memory fixtures,… - 95learner-orchestrator
Use when a learner wants a guided study session and needs help choosing among learning plans, Socratic tutoring, teach-back, retrieval practice, analogies, or critical review; route to one available learning skill at a time. - 85mastery
Mastery map for an env — the ideas behind its roadmap, each with a state that moves only on evidence of what the learner did. Triggers on "build my mastery map", "show my map", "what do I know in <env>", "why is <point> <state>". Learner-first; the tutor only asks. - 82new-topic
Start a new learning topic in an edu4self vault and put it on the dashboard. Use when the learner wants to learn something new and the current folder is an edu4self vault.