agentleFS
lightarktechlightarktech/founder-coding-skills/skills/context-hygiene/SKILL.md

context-hygiene

Skill1 starsChanged 4 months agoMITView on GitHub

Keep AI sessions short, rules in files, and context fresh. Use when sessions grow long, quality drifts, rules seem forgotten, or the founder asks why the AI "got dumber".

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 1

    Distinctive

    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:

    End of session = write the state down: Before closing mid-project, save status to a file or write a handoff (see handoff-contract).

    line 15
  2. Distinctive:

    No working session needed the phrase "as I told you before".

    line 19

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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

The most invisible failure mode for a non-technical founder: nothing crashed, but the AI slowly got worse. The cause is almost always context — a session dragging weeks of history, or rules that lived only in a chat that ended.

Rules

  • Rules live in files, never in chat. Anything meant to survive tomorrow ("from now on, always X") goes into a rules file (CLAUDE.md, a skill, a memory file) the moment it's decided. A rule that lives only in conversation dies with the conversation.
  • One ticket, one fresh session for building work (see vertical-tickets). Long-lived "everything" chats accumulate noise the AI re-reads on every turn — you pay more and get dumber answers (context rot). Our own logs: agents past ~250 turns were re-reading 100k+ tokens of history per turn, at premium prices, with visibly degrading judgment.
  • The founder's own thread is exempt. A founder's long strategy conversation is continuity, not waste — never nag them to close it. The cost lives on the AI's side: dispatch chores out, and suggest a single mid-thread compaction at a natural milestone if the session grows heavy — once, then drop it. Fresh-session discipline governs build work, not the founder's memory of their company.
  • Long chat you can't leave? Compact it. Summarize-and-continue (e.g. /compact) rebuilds the context and reloads the current rules. Even better when picking up old work: close and resume the session — full history kept, rules loaded fresh.
  • New rules don't reach old sessions. Sessions load rules at start. After changing the rules files, restart or compact any session that must follow them — don't assume it noticed.
  • End of session = write the state down. Before closing mid-project, save status to a file or write a handoff (see handoff-contract). "The AI will remember" is false; files remember.

Success criteria

No working session needed the phrase "as I told you before". Rule changes reached every active session the same day. Nothing important exists only inside a closed chat.

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.