mped-skills
CLAUDE.md11 starsChanged 3 months agoMITView on GitHub
In this repository18 other files it gives its agents.
- android-app-reverserSkill
- android-emulator-app-live-inspectSkill
- aula-pii-scanSkill
- de-ai-proseSkill
- de-ai-repoSkill
- drawio-figure-reviewerSkill
- explainer-videoSkill
- justfile-hygieneSkill
- model-raceSkill
- nix-scaffoldSkill
- phase1-prd-grillSkill
- phase2-backlog-snowballSkill
- phase3-backlog-ralphSkill
- readme-improverSkill
- refactor-smellerSkill
- scrub-for-publicSkill
- xterm-tmux-controlSkill
- zellij-controlSkill
The file
Show its rules and text
We split this CLAUDE.md into the rules it gives an agent
28 rules. About 0 are common, 26 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈26
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 2
Always run end-to-end tests before git commit: just e2e
- Distinctive: line 4
If changes are intentional, update expected output by reviewing test failures and regenerating snapshots
- Distinctive: line 6
Always let qa-test-runner sub-agent run before git commit.
- Distinctive: line 7
Always let mped-architect review the code-change before git commit.
- Distinctive: line 8
Run qa-test-runner and mped-architect in parallel
- Distinctive: line 12
Dont be sycophantic, I want to know about problems, gotchas and limitations.
- Distinctive: line 13
Dont be overly confident that the code is working - test it.
- Distinctive: line 14
Dont use emojis so much.
- Distinctive: line 17
Dont be eager to mention quickly-rotting number stats or reference line numbers.
- Distinctive: line 18
Dont publish commits, issues or PRs to GitHub without approval.
- Distinctive: line 22
NEVER claim success when something doesn't work: - Test thoroughly and report actual results
- Distinctive: line 23
When something is broken:
- Distinctive: line 26
Test that the fix actually works
- Distinctive: line 27
If you cannot fix it, clearly state why and what would be needed
- Distinctive: line 28
Incomplete work is failure: - Dont present partial solutions as complete
- Distinctive: line 29
Test end-to-end: - Verify the entire feature works, not just individual parts
- Distinctive: line 30
When asked to fix something, the job is not done until it actually works
- Distinctive: line 34
Dont make shell.nix file's shellHook spammy with verbose echos - only what is needed.
- Distinctive: line 35
NEVER add verbose echos to shell.nix shellHook
- Distinctive: line 36
If there is a flake.nix or shell.nix file present, use it.
- Distinctive: line 39
Place common commands (build, lint, test, run and so on) in Justfile as recipes.
- Distinctive: line 40
Run just recipes from within the nix-shell (if any)
- Distinctive: line 43
If an Electron app (drawio, chromium, vscode, etc.) flashes briefly then exits silently with code 0, it's almost certainly stale Singleton …
- Distinctive: line 48
Dont produce summary files as they go stale quickly - better to add to commit message.
- Distinctive: line 49
If there are stale outdatad irrelevant files, code or docs, then move to cruft/ folder.
- Distinctive: line 50
Never move backlog md tasks to cruft/ folder.
- Very specific: line 53
Never assume or override my git user name or email.
- Distinctive: line 54
Do not run backlog tasks in the background.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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.
Files
Development Workflow
- Always run end-to-end tests before git commit:
just e2e- E2E test failures BLOCK commits - you must fix the issues or update expected output
- If changes are intentional, update expected output by reviewing test failures and regenerating snapshots
- E2E tests verify the integrated system works, not just individual units
- Always let qa-test-runner sub-agent run before git commit. Fix what is reported by the sub-agent.
- Always let mped-architect review the code-change before git commit. Fix what is reported by sub-agent.
- Run qa-test-runner and mped-architect in parallel
Personal Guidelines
- Be self-critical. Point out assumptions and think of potential pitfalls.
- Dont be sycophantic, I want to know about problems, gotchas and limitations.
- Dont be overly confident that the code is working - test it.
- Dont use emojis so much.
- Believe in yourself and be optimistic. Dont give up when faced with difficult or complex problems.
- Seek to minimize complexity. Reduce accidental complexity.
- Dont be eager to mention quickly-rotting number stats or reference line numbers. Both go stale the moment the code changes. Prefer a name anchor, a grep recipe, or a list over a count.
- Dont publish commits, issues or PRs to GitHub without approval. Ask first, every time. Local commits are fine; anything that leaves the machine is not.
Critical Engineering Standards
- NEVER implement workarounds - Fix the root cause or clearly state you cannot fix it
- NEVER claim success when something doesn't work - Test thoroughly and report actual results
- When something is broken:
- Identify the exact root cause
- Fix the actual problem, not symptoms
- Test that the fix actually works
- If you cannot fix it, clearly state why and what would be needed
- Incomplete work is failure - Dont present partial solutions as complete
- Test end-to-end - Verify the entire feature works, not just individual parts
- When asked to fix something, the job is not done until it actually works
- Be honest about failures - Admit when you've failed to solve the problem properly
Nix Development
- Dont make shell.nix file's shellHook spammy with verbose echos - only what is needed.
- NEVER add verbose echos to shell.nix shellHook
- If there is a flake.nix or shell.nix file present, use it. Prefer flakes.
Justfile
- Place common commands (build, lint, test, run and so on) in Justfile as recipes.
- Run just recipes from within the nix-shell (if any)
Troubleshooting: Electron Apps (DrawIO, Chrome, etc.)
- If an Electron app (drawio, chromium, vscode, etc.) flashes briefly then exits silently with code 0, it's almost certainly stale Singleton IPC files. The app thinks another instance is running and tries to hand off to it, but the other instance is dead.
- Fix:
rm ~/.config/<app-name>/Singleton*(e.g.rm ~/.config/draw.io/Singleton*) - Also remove the stale socket if it exists (check the
SingletonSocketsymlink target under/tmp/)
Cruft and stale files
- Dont produce summary files as they go stale quickly - better to add to commit message.
- If there are stale outdatad irrelevant files, code or docs, then move to cruft/ folder.
- Never move backlog md tasks to cruft/ folder.
- linux-utils is a script, already in $PATH, that shows other useful tools related to git, gitlab, backlog, etc. Whenever gitlab or backlog is referenced, run linux-utils to learn what is available.
- For backlog tasks, only use backlog tool - never edit backlog md files directly.
- Never assume or override my git user name or email; gitconfig already has rules for it. Use the repo's existing config — do not pass
git -c user.name=...orgit -c user.email=...on commits. - Do not run backlog tasks in the background. Look at their stderr and stdout to check.
- ALWAYS use
--plainflag forbacklog task view,backlog board, and other commands that have interactive TUI modes. Without--plain, these commands launch a TUI that waits for keyboard input and will hang indefinitely in non-interactive contexts.
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
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.

