FR

frontend-mix-validate

Performs automated validation tasks on integrated apps to ensure code quality and build stability.

Install

mkdir -p .claude/skills/frontend-mix-validate && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/17450" && unzip -o skill.zip -d .claude/skills/frontend-mix-validate && rm skill.zip

Installs to .claude/skills/frontend-mix-validate

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Run the full validation suite (install, type-check, lint, build, tests) on a freshly integrated app and fix anything broken. Run this skill in a Claude Code session driving Sonnet - validation is grind work, Sonnet is the right tool. Two-attempt cap per step; failures are recorded for the fix-validation escalation step, not papered over. Use after the frontend-mix-integrate session finishes.
394 charsno explicit “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Intermediate

Key capabilities

  • Run package installation using `bun install`, `npm install`, or `pnpm install`
  • Perform type checking with `tsc --noEmit` or `bun run typecheck`
  • Execute linting commands like `bun run lint` or `eslint .`
  • Build the application using `bun run build` or framework equivalent
  • Run tests if a test script exists

How it works

This skill runs a full validation suite on an integrated app, including installation, type-checking, linting, building, and testing, with a two-attempt cap per step.

Inputs & outputs

You give it
Path to an integration summary markdown file
You get back
A validation summary or validation issues markdown file, detailing success or failures

When to use frontend-mix-validate

  • Validating new frontend integrations
  • Automating build and lint checks
  • Debugging integration failures

About this skill

Frontend-Mix · Validate

You are the validation step of a manual mixed-provider build. Sonnet handles this - it's grind work, not judgment. Two attempts per check, then record the failure. Never paper over real bugs.

What to do

  1. Use the Read tool to open $ARGUMENTS end-to-end. Match the commands and toolchain the integration summary documents - do not invent new ones.

  2. The filename in $ARGUMENTS carries your run-name. Strip the directory and the -integration-summary.md suffix. You'll use it to name your output file.

  3. If $ARGUMENTS is empty, ask the user for the integration summary path. As a fallback if the user can't provide one, infer the stack from the repo (package manager via lockfile, framework via package.json).

Steps (each capped at 2 attempts)

  1. Install: bun install (or npm install / pnpm install based on lockfile).
  2. Type check (tsc --noEmit, bun run typecheck, etc).
  3. Lint (bun run lint, eslint ., etc).
  4. Build: bun run build (or framework equivalent).
  5. Tests if a test script exists.

For each step:

  • Attempt 1: run the command. If it passes, move on.
  • Attempt 2: if it failed, try one targeted fix (read the error, edit the offending file, re-run). If it still fails, STOP retrying that step and record the failure.

Failure rules (no shortcuts)

  • Do NOT paper over real bugs. A failing type check on any is a bug.
  • Do NOT delete tests to make them pass.
  • Do NOT add // @ts-ignore, // eslint-disable, weaken types, or skip steps.
  • If you find yourself wanting to do any of that, stop and record the failure.

Output - exactly ONE of these two files

Both go to .claude/artifacts/ with the <run-name> prefix. Create the directory if it doesn't exist.

Case A - All steps passed cleanly

Write .claude/artifacts/<run-name>-validation-summary.md containing:

  • Each command run with its exit code
  • One-line confirmation: "All checks passed."

Case B - One or more steps failed after 2 attempts

Write .claude/artifacts/<run-name>-validation-issues.md containing, for each failure:

  • Step name (e.g. "type check")
  • Exact command that failed
  • Truncated error output (last 30 lines)
  • File(s) most likely responsible (your best guess)
  • One-sentence hypothesis for the root cause

Then STOP. The next session (fix-validation, run with Opus) reads this file and addresses each failure with judgment.

After validating

Tell the user the absolute path to whichever file you wrote, and the next step:

If you wrote <run-name>-validation-summary.md (clean):
  Next: invoke /frontend-mix-smoke with the integration-summary path and the validation-summary path.

If you wrote <run-name>-validation-issues.md (failures recorded):
  Next: switch back to Opus and invoke /frontend-mix-fix-validation with the validation-issues path and the plan path.

When not to use it

  • When the task is not to validate a freshly integrated app
  • When attempting to paper over real bugs by ignoring errors or deleting tests
  • When more than two attempts have been made for a step

Limitations

  • Two-attempt cap per step for fixes
  • Does not paper over real bugs or weaken types
  • Requires an integration summary path as input

How it compares

This skill provides a structured, automated validation process with strict failure rules and a two-attempt limit, ensuring thorough post-integration cleanup compared to manual checks.

Compared to similar skills

frontend-mix-validate side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
frontend-mix-validate (this skill)01moNo flagsIntermediate
verification-loop03moReviewIntermediate
dependency-upgrade264moReviewIntermediate
validate-typescript59moReviewBeginner

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry