Skip to content
Agentmart
Submit a tool

learn

Records a new rule in memory, project or user rules (or the plugin source, if set).

Skill by stage2dev

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/stage2dev-learn

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Records a new rule in memory, project or user rules (or the plugin source, if set). Use for "remember this", "from now on", "new rule", "never again", "guarde na memória", "a partir de agora", "regra nova", "nunca mais".

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/learn

Tools it pre-approves 4

From allowed-tools in SKILL.md: the agent may use these without asking you each time.

  • Bash(sh
  • "${CLAUDE_PLUGIN_ROOT}/scripts/py"
  • scripts/guarda.py
  • info)

What's inside

Skill name
learn
Description
Records a new rule in memory, project or user rules (or the plugin source, if set). Use for "remember this", "from now on", "new rule", "never again", "guarde na memória", "a partir de agora", "regra nova", "nunca mais".
License
MIT
Allowed tools
Bash(sh, "${CLAUDE_PLUGIN_ROOT}/scripts/py", scripts/guarda.py, info)
Instructions
35 lines in SKILL.md · view on GitHub

Before installing

  • Pre-approves some shell commands
  • Instructions read files in your home folder (~/.ssh, ~/.aws…)

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Learn - memory + rules

Context: !`sh "${CLAUDE_PLUGIN_ROOT}/scripts/py" scripts/guarda.py info`

Rule: $ARGUMENTS

The JSON above is the context. Write to the user in its `language`, honoring `reply_style`; final answer short. Use marker names from `names`, never literal. The owner already asked: do not ask "may I save?". Ask only if the rule is ambiguous (one question, recommended first).

## Steps

0. **Type.** One-off correction or fact -> memory only. A working rule that holds from now on -> memory + a rule file (or the plugin, if `plugin_source`).
1. **Memory** (only if the project has a Claude memory folder). Look for a duplicate first (`grep -ril` in the project's memory folder). Exists: update it. Else create a `feedback` memory with the rule, `**Why:**` (what happened or what the owner said) and `**How to apply:**` (when it holds). Link related ones with `[[other-memory]]`. Add 1 line to the `MEMORY.md` index.
2. **Where the rule lives.** Pick ONE, in this order of preference:
   - **Hook**: mandatory and detectable in the command or call (zero context cost). Plugin change: only with `plugin_source` (step 3).
   - **Project config** (`.tune-up/config.json`, `bash_rules`): a shell lock, only for that project.
   - **Project profile** (section "Rules" of `.tune-up/project.md`): a rule for that project.
   - **User rules** (`~/.config/tune-up/rules.md`): a rule for every project of the owner.
   - **Memory only**: a fact with no repeatable action.
3. **Apply.** `info.plugin_source` is null (the usual case): write the rule at the project or user level as above, and tell the owner in 1 line that changing the core needs a checkout of the source (`plugin_source`). NEVER write into `~/.claude/plugins/cache` or any copy of the plugin. `info.plugin_source` is set: edit the plugin files in that git checkout. New hook: new test in `tests/`. New or changed skill: an eval case in `evals/` if triggering matters. Keep text short; description <= 250 characters; skill body <= 150 lines; code file <= 450 lines.
4. **Verify** (only when the plugin was edited). In the checkout: `python3 -m unittest discover -s tests` (`python` where `python3` is missing) and `claude plugin validate <plugin_source> --strict`. Show the real result. Failed: fix it or say you did not.
5. **Version** (only when the plugin was edited). Add a `CHANGELOG.md` entry and bump `version` (patch) in `plugin.json`.
6. **Commit** (only when the plugin was edited), in the `plugin_source` repository, only the touched files, by name: `git -C <plugin_source> add <files> && git -C <plugin_source> commit -m "<summary>"`. Only if the owner's request allows commits.
7. **Reload.** Plugin edited: ask the owner to run `/reload-plugins` (the current session does not reread by itself).

## Rules

- A secret never goes into memory, config or a commit.
- A rule that contradicts a saved one: show both, say which wins (the owner's most recent) and update the old one.
- Do not raise fixed cost needlessly: if a hook solves it, write no skill text.

## Answer

Where it was saved (memory + the file or plugin checkout), what was tested and passed (if the plugin was edited), and `names.need_you` run `/reload-plugins` (if it was).

From stage2dev/tune-up, MIT.

More from stage2dev/tune-up

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 20, 2026 (9 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from stage2dev

Cite this page

learn skill. Agent Mart. https://agentmart.dev/skills/stage2dev-learn (last updated 2026-10-24)