drafael/coding-harness / skills/teach
teach
VerifiedTeach repository-backed code and engineering work through a progressive, evidence-grounded conversation. Manual invocation only; use when the user explicitly invokes teach to understand a change, subsystem, architecture, runtime behavior, or evidenced rationale. Read-only; do not modify files.
README.md
View on GitHub# Teach skill
Teaches repository-backed code and engineering work through a progressive, evidence-grounded conversation. It explains what something is and how it works. When rationale matters, it uses available evidence to explain why it was built that way. It does not change the repository.
## Activation
Manual only. Use the host's explicit skill-invocation mechanism. In Pi, run:
```text
/skill:teach
```
If a host does not enforce manual-invocation metadata, load the skill explicitly before asking it to teach. Ordinary code explanations remain with `explain-code`.
Example requests after invocation:
- "Teach me how request retries work in this service."
- "Help me understand this diff and why the state transition is necessary."
- "Teach me the event-processing subsystem one layer at a time."
- "Show me how this component fits into the application architecture."
## Behavior
- Infers the target, purpose, and useful starting depth from the conversation.
- Establishes mechanics from repository evidence and investigates rationale only when relevant.
- Separates observed behavior, documented rationale, inference, and unknowns.
- Starts with the smallest complete explanation and adds depth as the person responds.
- Uses code excerpts and diagrams only when they materially improve understanding.
- Remains read-only and routes requests for changes to another workflow.
When installed and loadable, `teach` can use `explain-code` for mechanics, `clarify` for rationale, and `unslop` for its final prose pass. These integrations are optional. The core includes local fallbacks and does not fail when companion skills are absent.
## Boundaries
- `explain-code` owns ordinary walkthroughs and current-code mental models.
- `clarify` owns standalone expectation mismatches and historical-rationale investigations.
- `code-review` owns defects, risks, and architectural criticism.
- `brainstorm` owns future designs and alternatives.
- `technical-writing` owns substantial standalone documents.
- `teach` owns explicitly invoked progressive learning about existing repository work.
## Structure
- `SKILL.md`: self-contained teaching, evidence, pacing, and visual workflow.
- `README.md`: activation, examples, boundaries, structure, and provenance.
- `LICENSE`: retained upstream MIT notice.
The skill does not require a named model, provider, tool, command syntax, subagent type, runtime, or agent harness. Host-specific invocation is documented only as an example.
## Origin and license
This local, vendor-neutral adaptation is inspired by [Cursor pstack `teach`](https://github.com/cursor/plugins/blob/fd6dd6f7276956a532bb78a748a8d2818b6eb5f4/pstack/skills/teach/SKILL.md), written by Lauren Tan.
The upstream skill is distributed under the MIT License. See `LICENSE` for the retained notice.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |