sanand0/scripts/code

Claude Code and agent skill by sanand0/scripts

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

ALWAYS follow this style when writing Python / JavaScript / HTML / CSS code

Install

npx vskill@latest install sanand0/scripts/code

SKILL.md

  • Inspect relevant code, tests, config, sample input/output before editing. Don't assume structure or behavior you can check directly
  • Minimize new code and changes. Prefer the first working option:
  1. Skip if unnecessary
  2. Reuse existing code
  3. Use native platform features
  4. Write it one line if that is clear and correct
  5. Use standard library
  6. Use existing library
  7. Use popular, modern, minimal, fast library
  8. Write the minimum code that works
  • Prefer deletion over addition, boring over clever, 1 file over many. Minimize abstractions, scaffolding, or dependencies
  • Prefer data over code: structures beat conditionals. Prefer config.{json|yaml|toml|...} if >= 30 lines
  • DRY: Use helpers for logic repeated 3+ times, precompute shared intermediates
  • Keep happy path linear and obvious. Write flow first, then fill in code
  • Early returns fail fast and reduce nesting. Skip defensive fallbacks, existence checks, ... unless essential
  • Never skip: validation at trust boundaries (user input, files, APIs), error handling that prevents data loss, security, accessibility
  • Change existing code minimally. Retain existing comments. Follow existing style.

Preserve behavior outside requested changes. Ask before materially expanding scope or touching unrelated files/config

  • Make scripts re-startable if interrupted. Inspect state first for unexpected changes

Docs:

  • Use type hints and docstrings (document contracts and surprises, not mechanics)
  • Comment non-obvious stuff that'll trip future maintainers: why, why not alternatives, pitfalls, invariants, input/output shape, ...
  • Mark deliberate shortcuts with a known limit: shortcut: <limit>; <upgrade path> (e.g. # shortcut: O(n²) scan; index by id past ~10k rows)

Tests:

  • When tests exist, or writing new code, add and run tests first
  • Use existing commands and environment - don't reconstruct dependencies ad hoc
  • Run minimal existing tests for changed behavior. Extend relevant existing tests before creating new test/file
  • Lint/format only files you changed. Don't auto-format legacy code unless asked
  • Test final outputs, not just the source / intermediates
  • Test user-visible front-end changes with:
  • DOM/source checks for behavior: interactions, navigation, overlaps, cut-off elements, readability, colour/font size count, ...
  • Screenshots for appearance changes: responsive layout, overlaps, contrast & visibility, visual impact, ...
  • Lighthouse audit via Chrome DevTools MCP else npx -y lighthouse@latest: when semantics, controls, colors, performance changes
  • Keep tests fast
  • Never say "verified" without evidence. List changes, validations with results, and remaining risks/unknowns

Ops:

  • Log status & progress for long tasks (>5s). Log _before_ action. Flush logs

Bug fixes:

  • Fix the root cause, not just the reported symptom. Check callers before patching

Python

Avoid python or python3. Prefer uv run script.py for PEP 723 scripts and uv run --with pkg ... for ad-hoc code/tests. Tests importing a PEP 723 script must use the repo test environment or explicitly include its dependencies.

Avoid requirements.txt. Unless pyproject.toml is present, add dependencies as PEP 723 metadata:

py
#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.14"
# dependencies = ["scipy>=1.10", "httpx"]
# ///

Prefer when really needed:

  • typer|click > argparse
  • httpx2 > httpx > requests
  • lxml > xml
  • duckdb > pandas > csv
  • tenacity for retries

Web

Prefer modern HTML/CSS/platform features over custom JavaScript or libraries.

  • Images: responsive srcset/sizes; loading="lazy" for offscreen images, but never likely LCP images; use fetchpriority="high" for important/LCP images
  • Forms: semantic input types, autocomplete=, inputmode=, enterkeyhint=, list=, autocapitalize=, spellcheck=, form=
  • UI: <dialog>, popover, popovertarget=, commandfor=/command=, inert, <details name="">, closedby=
  • Semantics: prefer native elements such as <search>, <meter>, <progress>, <output>, <data>, <time> over custom equivalents
  • Media: <picture>, srcset=, preload=, poster=, playsinline, <track>
  • CSS: prefer container queries for component responsiveness; logical properties, clamp(), text-wrap: balance|pretty, color-scheme, light-dark(), color-mix(), oklch() where useful
  • Motion: prefer CSS transitions/animations over JavaScript; respect prefers-reduced-motion
  • Accessibility: preserve native keyboard/focus behavior; use :focus-visible; prefer native HTML semantics over ARIA/custom JavaScript
  • Validation: prefer native constraint validation and :user-valid/:user-invalid
  • Prefer Baseline Widely Available features. For newer features, feature-detect / use @supports and degrade gracefully; avoid polyfills/dependencies unless required.
  • For unfamiliar/new web-platform features or uncertain browser support, consult npx -y modern-web-guidance@latest search "<task>" and retrieve only the relevant guide.

JavaScript

Preferred JS style:

  • Hyphenated HTML class/ID names (id="user-id" not id="userId")
  • Use modern JavaScript and ES modules: ?., ??, destructuring, spread, implicit returns (=> over => { return })
  • No TypeScript, but // @ts-check. .d.ts is OK for packages
  • With fetch(), check response.ok before consuming the response unless non-2xx responses are intentionally handled
  • Show loading/pending state for user-visible async operations
  • Handle expected/recoverable errors where they occur; let unexpected errors propagate to a top-level handler
  • Prefer textContent/DOM APIs for untrusted content; never interpolate untrusted data into innerHTML

For browser debugging, use the devtools skill. For single-page HTML, prefer file:// when a server isn't required.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/sanand0/scripts/code/badge)](https://verified-skill.com/skills/sanand0/scripts/code)

Something wrong with this skill or its page? Tell us.