Skip to content
Development
Agent

learn-this

Loaded when the user invokes `/learn-this`, says "learn this" / "remember this", or accepts your auto-detect offer to capture a correction.

From plugin
think-like-me
132 skills2 agents
Install
$ npx -y skills add kakha13/think-like-me --agent claude-code

How it fires

How this agent gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.

Context preview

The summary Claude sees to decide when to auto-load this agent.

Loaded when the user invokes `/learn-this`, says "learn this" / "remember this", or accepts your auto-detect offer to capture a correction.

Agent definition

learn-this.md

Learn this — from a correction

Loaded when the user invokes `/learn-this`, says "learn this" / "remember this", or accepts your auto-detect offer to capture a correction.

Inputs

  • **The correction.** The user's most recent message that corrects a prior response.
  • **The corrected response.** The assistant message the correction replaced.
  • **Task context.** Preceding messages that set up the task being worked on.

Procedure

1. Analyze

Extract these fields from the conversation:

  • **Task:** what was the user trying to accomplish?
  • **Attempt:** what did you propose?
  • **Correction:** what did the user say was wrong?
  • **Fix:** what's the correct approach, per the user's correction?
  • **Principle:** what's the generalizable rule underneath this specific correction? Do not save a rule about "this specific endpoint" — generalize to the pattern.
  • **Mechanism (the why):** what makes the wrong approach fail? If the user didn't say, infer from the correction language and your understanding of the domain. If you can't infer confidently, ask.

2. Draft the rule

Fill in the canonical block. `Don't` is often set here — it's literally what you just did wrong.

### <Short, searchable title>
**When:** <domain> + <feature keywords> + <specific context>
**Do:** <the fix, generalized>
**Don't:** <what you reached for that was wrong>
**Why:** <the mechanism>
**Example:** <optional — include only if non-obvious>

3. Pick a category

Match to one of: `mobile`, `backend`, `ui`, `general`. If none fits, propose creating a new one (e.g. `devops`, `data`). Confirm the new category name with the user before writing.

4. Show the draft

Present in one message:

> **Draft rule → `rules/<category>.md`** > > [full block verbatim] > > Save? (`yes` / `edit` / `skip`)

5. Handle response

  • `yes` → dual-write per `SKILL.md`'s "Saving a rule" section.
  • `edit` → ask what to change, revise, re-show. Repeat until `yes` or `skip`.
  • `skip` → drop silently. Do not re-prompt.
  • Anything else / silence → drop silently.

Quality bar

  • **Generalize.** A rule about "the `/orders/:uuid/sync` endpoint returned 500" is too narrow. Generalize to the pattern that caused the 500. If you cannot generalize, the correction might not be rule material — just fix and move on.
  • **`Why` is mandatory.** No rule without a mechanism. If you can't state one, ask the user.
  • **Title is a lookup key.** 3–8 words. Start with the domain or feature, not "Don't" or "Always".
Read more
Ships withthink-like-me

A personal Claude Code skill that stores your curated engineering rules and actively consults them before Claude drafts code. Complements Claude Code's passive auto-memory: memory surfaces facts when relevant, rules shape the approach up front.

Get the whole plugin
Stats
13
Stars
0
Forks
Maintained
Maintenance
HTML
Language
5mo ago
Last commit
5mo ago
Created

Repo: kakha13/think-like-me

Other agents on think-like-me.