agentleFS
michielvandermeermichielvandermeer/skills/skills/learn-habits/SKILL.md

learn-habits

Skill9 starsChanged 19 days agoMITView on GitHub

Rebuild a person's Grilling profile from the Habits in their past grilling sessions.

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Before writing, read technical-writing/SKILL.md and follow it — /technical-writing does not load as a skill.

    line 36
  2. Distinctive:

    Done: when that commit is made.

    line 61

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

A Grilling profile holds one person's Habits in this repo, at .agents/refs/profiles/<name>.md, so grilling can predict which option each person would pick. It learns only from that person's grilling sessions in this repo; the same person may choose differently elsewhere. See ADR-0063.

Whose profile

The person is whoever runs this session: git config user.email. When that is empty, stop and say that no profile can be found or written without it.

One person may commit under a different email on each machine, so a profile's Email: line lists every one of them. Their profile is the file in .agents/refs/profiles/ whose Email: line lists their email. None lists it → their profile is the file named after git config user.name in kebab case (ana-silva.md): add the email to its Email: line, or create it when it does not exist. Write only this person's profile; a teammate's file is theirs to write from their own sessions.

Rebuild

Typed by the user, never loaded by a grilling skill. Rebuild the profile from scratch over a window: the last 100 grilling sessions in this repo, or the window the argument names. A rebuild replaces the whole file except its Email: line.

  1. Find the sessions. Read the session logs of every host on this machine, not only this one. Look up where each keeps them; today Claude Code keeps ~/.claude/projects/<cwd with / as ->/<session>.jsonl (skip /subagents/), and Grok keeps ~/.grok/sessions/<url-encoded cwd>/<session>/, where only a session whose prompt_context.json says "audience": "primary" is top-level. A worktree's sessions belong to its repo. A grilling session is one where an assistant message holds a ## Round N heading and an option marked ← recommended, whichever skill ran it; that text inside a file the session merely read is not a Round.
  2. Read them in batches. Dispatch skills:explorer sub-agents, about 15 sessions each, in parallel. Each returns one record per answered Question: session date, the kind of question in a few plain words, the options, which one was recommended, what the person picked or wrote, any *Your habit:* line and the Habit it named, and the person's stated reason, paraphrased. Carry their records, never the logs.
  3. Build the profile from the records under Habit rules.

Habit rules

An override is an answer other than the recommended option, the person's own words included. An answer that corrects a fact about the domain, or what the person meant, is a missing fact rather than a Habit — leave it to the glossary and the ADRs.

  • A pattern is overrides that go the same way on the same kind of question. State it as a rule that names that kind of question: "include every related area when asked where a change should stop", never a bare direction such as "prefers wider scope" — two Habits going opposite ways on different kinds of question are both true.
  • A Habit is a pattern seen in at least 3 sessions. A pattern seen in 1 or 2 stays under ## Patterns until a third session shows it.
  • Its record is every question it applied to, last 10 kept, oldest first: ✓ when the person picked what the Habit points to, ✗ when they went the other way. A question where a *Your habit:* line moved the recommendation counts, accepted or not.
  • Drop a Habit when its last 10 hold more ✗ than ✓.
  • Merge Habits that overlap. Every grilling session reads every profile, so keep the file short. Paraphrase; quote nothing.

Before writing, read technical-writing/SKILL.md and follow it — /technical-writing does not load as a skill. Teammates read the profile too.

Profile shape

# Grilling profile: Ana Silva

Email: ana@example.com, ana.silva@client.example
Sessions: 27, from 2026-08-28 to 2026-10-09
Picked the recommendation: 131 of 180 answers (73%)

## Habits

- **Include every related area when asked where a change should stop.** 9 sessions, last 2026-10-09. Last 10: ✓✓✗✓✓✓✓✓✗✓
- **Leave out a safeguard for a case that has no users or data yet.** 6 sessions, last 2026-10-07. Last 10: ✓✓✓✓✓✓

## Patterns

- Name a new setting the way the nearest existing one is named. 2 sessions, last 2026-10-02.

Commit

Stage the profile by name and commit it on its own — learn-habits: <first name>'s Grilling profile, naming the window it rebuilt from.

Done when that commit is made.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.