mirror of https://github.com/alirezarezvani/claude-skills.git synced 2026-09-07 08:26:02 +00:00

fix(docs): render orphan sub-skills (recover 79 missing skill pages)

generate-docs.py had a longstanding bug: the rendering loop only iterated
top-level skills and only rendered their direct children. Sub-skills whose
parent is a plugin folder (not a top-level skill at <domain>/skills/<name>/)
were silently dropped.

Affected plugins (standalone-only, no bundled mirror at <domain>/skills/):
- executive-mentor (1 index + 5 sub-skills)
- agenthub (1 index + 7 sub-skills)
- autoresearch-agent (1 index + 5 sub-skills)
- playwright-pro (1 index + 9 sub-skills)
- self-improving-agent (1 index + 5 sub-skills)
- c-level-agents (1 index + 17 sub-skills — the new /cs:* commands)
- llm-wiki (1 index + sub-skills)
- behuman, code-tour, demo-video, helm-chart-builder, karpathy-coder,
  llm-cost-optimizer, prompt-governance, statistical-analyst, terraform-patterns,
  data-quality-auditor, docker-development (single-skill plugins)

Total: 79 sub-skills + 12 plugin-index skills = 91 pages were being dropped.
(Some plugins like behuman are single-skill so only their index is dropped.)

The bug: rendering loop at line 414 only handled `for skill in top_level`,
then for each top-level found `children = [s for s in sub_skills if
s["parent"] == skill["name"]]`. Plugins where the SKILL.md lives only at
<plugin>/skills/<plugin>/SKILL.md don't appear in top_level (their detection
puts them in sub_skills with parent=themselves), so their children were
orphaned.

The fix: after the existing top-level loop, render orphan sub-skills grouped
by their plugin parent. Index sub-skill (named same as parent) renders as
<parent>.md; other children render as <parent>-<child>.md. This matches the
URL convention already in use (e.g., executive-mentor-challenge.md), so
existing SEO equity is preserved.

Result: skill pages generated 193 → 272 (+79 recovered). Total docs pages
280 → 359. mkdocs build succeeds.

Verified:
- All 12 previously-dropped plugins render their index page
- All 79 previously-dropped sub-skills render their detail pages
- URL convention preserved (executive-mentor-challenge.md, agenthub-board.md,
  playwright-pro-coverage.md, etc.)
- karpathy diff_surgeon: 0 findings

After dev → main release: GitHub Pages redeploys with the recovered 79 pages.
The docs site finally has 1:1 correspondence between SKILL.md files in the
repo and pages on the site.

https://claude.ai/code/session_012WtZMm5NJHqkYoRqA9fHMN

2026-05-13 12:49:00 +00:00

5.8 KiB

Raw Permalink Blame History

title	description
Karpathy Coder — Active Coding Discipline — Agent Skill for Codex & OpenClaw	Use when writing, reviewing, or committing code to enforce Karpathy's 4 coding principles — surface assumptions before coding, keep it simple, make. Agent skill for Claude Code, Codex CLI, Gemini CLI, OpenClaw.

Karpathy Coder — Active Coding Discipline

:material-rocket-launch: Engineering - POWERFUL :material-identifier: `karpathy-coder` :material-github: Source

Install: claude /plugin install engineering-advanced-skills

Derived from Andrej Karpathy's observations on LLM coding pitfalls. This is not just guidelines — it ships Python tools that detect violations, a review agent, a slash command, and a pre-commit hook.

"The models make wrong assumptions on your behalf and just run along with them without checking. They don't manage their confusion, don't seek clarifications, don't surface inconsistencies, don't present tradeoffs, don't push back when they should."

"They really like to overcomplicate code and APIs, bloat abstractions, don't clean up dead code... implement a bloated construction over 1000 lines when 100 would do."

"LLMs are exceptionally good at looping until they meet specific goals... Don't tell it what to do, give it success criteria and watch it go."

— Andrej Karpathy

The four principles

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

State assumptions explicitly. If uncertain, ask.
If multiple interpretations exist, present them — don't pick silently.
If a simpler approach exists, say so. Push back when warranted.
If something is unclear, stop. Name what's confusing. Ask.

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

No features beyond what was asked.
No abstractions for single-use code.
No "flexibility" or "configurability" that wasn't requested.
No error handling for impossible scenarios.
If you write 200 lines and it could be 50, rewrite it.

The test: Would a senior engineer say this is overcomplicated? If yes, simplify.

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

Don't "improve" adjacent code, comments, or formatting.
Don't refactor things that aren't broken.
Match existing style, even if you'd do it differently.
If you notice unrelated dead code, mention it — don't delete it.
Remove imports/variables/functions that YOUR changes made unused.
Don't remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user's request.

4. Goal-Driven Execution

Define success criteria. Loop until verified.

Instead of...	Transform to...
"Add validation"	"Write tests for invalid inputs, then make them pass"
"Fix the bug"	"Write a test that reproduces it, then make it pass"
"Refactor X"	"Ensure tests pass before and after"

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Slash command

/karpathy-check — Run the full 4-principle review on your staged changes.

Python tools (`scripts/`)

All tools are stdlib-only. Run with --help.

Script	What it detects
`complexity_checker.py`	Over-engineering: too many classes, deep nesting, high cyclomatic complexity, unused params, premature abstractions
`diff_surgeon.py`	Diff noise: lines that don't trace to the stated goal — comment changes, style drift, drive-by refactors
`assumption_linter.py`	Hidden assumptions in a plan: unasked features, missing clarifications, silent interpretation choices
`goal_verifier.py`	Weak success criteria: vague plans without verifiable checks, missing test assertions

Sub-agent

karpathy-reviewer — Runs all 4 principles against a diff. Dispatched by /karpathy-check or manually before committing.

Pre-commit hook

hooks/karpathy-gate.sh — runs complexity_checker.py and diff_surgeon.py on staged files. Warns (non-blocking) when violations are found. Wire it via .claude/settings.json or Husky.

References

references/karpathy-principles.md — the source quotes, deeper context, when to relax each principle
references/anti-patterns.md — 10+ before/after examples across Python, TypeScript, and shell
references/enforcement-patterns.md — how to wire hooks, CI integration, team adoption

When to relax

These principles bias toward caution over speed. For trivial tasks (typo fixes, obvious one-liners), use judgment. The principles matter most on:

Non-trivial implementations (>20 lines changed)
Code you don't fully understand
Multi-step tasks with unclear requirements
Anything that will be reviewed by humans

Cross-tool compatibility

Installs via plugin for Claude Code. For other tools, copy the principles into your schema file:

Tool	Schema file
Claude Code	`CLAUDE.md` (auto-loaded by plugin)
Codex CLI	`AGENTS.md`
Cursor	`AGENTS.md` or `.cursorrules`
Antigravity / OpenCode / Gemini CLI	`AGENTS.md`

Related skills (chains via `context: fork`)

self-eval — honest quality scoring after completing work
code-reviewer — broader code review; karpathy-coder focuses on the 4 LLM-specific pitfalls
llm-wiki — compound knowledge; karpathy-coder ensures you don't overcomplicate while building it

5.8 KiB Raw Permalink Blame History