⚡ Activation

💭 in your head · nothing to type

Bring to mind your project — the one you've been working in throughout the course.

Don't look at it as a "finished thing". Picture a workshop after six months of work. The tools lie where they were last dropped. The bench is covered in shavings from three different jobs, and in the corner sits a box labelled "sort out later". You can still work here. But every new job starts with hunting for the right wrench.

Which corner of your workshop do you keep walking around because it has needed tidying for a long time?

👁 Reflection

💭 in your head · nothing to type

Cleanup usually gets postponed not out of laziness. It's unclear where to start, and starting with everything at once is scary: what if something breaks?

Don't think of it as a "spring cleaning". Picture a car inspection before a long trip. The mechanic doesn't rebuild the engine. He goes down a list — oil, brakes, tyres, lights — and fixes whatever makes the trip unsafe. The rest he notes for next time.

Consider this: you already have a partner who can read your whole project in minutes. What stopped you from simply asking it to take a look?

💡 Concept

What "Pendel" is

Pendel is Russian slang for "a kick in the pants". In this practice you ask your agent to run a hygiene audit of your project. Then you fix three items from its report. Nothing to install: the agent you already work with is enough — Claude Code, Codex or any other that reads your project files.

The idea is simple: throughout the course you kept adding things to the project — a rules file, memory, hooks, skills. Pendel is the first time you look at the project as a whole and ask: what here gets in the way of working further?

Six audit zones

Zone What we look for Link to the course
1. File size and monoliths The longest files in the project; files that mix several responsibilities (routing, logic, markup in one place) A long file is a long context (05/u1)
2. Inline styles and mixed layers Styles and scripts inside HTML or templates instead of separate files Things kept separately are easier for the agent to edit
3. Clutter in agent settings Stale permissions, duplicates, absolute paths to other folders, keys and tokens in the config MCP and hooks config (07/u2, 07/u3)
4. The agent's rules file CLAUDE.md or AGENTS.md: length, contradictions, outdated instructions Aim for about 200 lines (05/u3)
5. Memory between sessions Is there a file with the project state: what's done, what's left, what problems exist Context compression (05/u4)
6. Hooks Is there at least one automatic check; is it clear what the existing ones do Hooks and their security (07/u3)

If your agent doesn't support a zone (hooks, for example), it should say so in the report. That's a finding too.

The audit prompt

Copy it whole and send it to your agent at the root of your project: