← Browse

@hcompai/hai-qa-via-cli

A

Run an autonomous QA review of a web UI by invoking the local `qa-cli` command via Bash. Trigger when the user asks to QA a URL, check a deployed site for accessibility, usability, or correctness issues, get a structured review of a page, or run a quick visual sanity check on a live page (e.g. "what color is the CTA?").

skillclaude

Install

agr install @hcompai/hai-qa-via-cli --target claude

Writes 1 file into .claude/skills/, pinned to git-e94b47e8.

  • .claude/skills/hai-qa-via-cli/SKILL.md

Document


name: hai-qa-via-cli description: > Run an autonomous QA review of a web UI by invoking the local qa-cli command via Bash. Trigger when the user asks to QA a URL, check a deployed site for accessibility, usability, or correctness issues, get a structured review of a page, or run a quick visual sanity check on a live page (e.g. "what color is the CTA?"). metadata: version: "1.0.1" author: hcompai repository: https://github.com/hcompai/computer-use-agents-demos/skills

Web QA via the hai-agents CLI

When the user asks you to QA a web UI / review a deployed page / check a URL for issues:

1. Pick the right subcommand

  • qa-cli review — full agent loop (~30–90s). Returns a structured ReviewResult JSON with verdict, summary, findings (each with severity, area, issue, suggestion), and steps_taken. Use this for any real review.
  • qa-cli visual — one-shot visual question, returns a short free-text answer. Use for quick "what color is X" / "does the layout look broken" sanity checks.

2. Invoke

uv run qa-cli review --url <url> --instruction "<what to focus on>"
uv run qa-cli visual --url <url> --question "<single question>"

The CLI logs progress to stderr; the result (JSON for review, text for visual) goes to stdout. Pipe to jq if you want to filter.

3. Surface the result

For review:

  • Lead with the verdict and the headline summary.
  • List findings ordered as returned (already high-severity-first). Show each as [severity] area — issue and include the suggestion when actionable.
  • If verdict == "fail", make it clear what blocked completion.

For visual: just relay the answer.

Notes

  • Prerequisite: the qa-cli command comes from the computer-use-agents-demos repo. If uv run qa-cli ... errors with "command not found", that repo isn't installed in the current working directory — tell the user to clone it and run uv sync before retrying, or fall back to whichever QA tool is already on PATH.
  • Requires HAI_API_KEY in .env. On a missing key the CLI exits with error: HAI_API_KEY is not set... — relay that to the user verbatim; don't retry.
  • review is expensive (one cloud session per call). Don't run multiple reviews in parallel for the same task — call it once with a focused --instruction.

Examples

User: "Please QA https://news.ycombinator.com for accessibility." You: Run uv run qa-cli review --url https://news.ycombinator.com --instruction "focus on accessibility — alt text, form labels, keyboard nav, contrast". Parse JSON. Surface findings.

User: "What color is the main CTA on https://example.com?" You: Run uv run qa-cli visual --url https://example.com --question "what color is the main call-to-action button?". Return the answer.

Trustgrade A

  • passBody integrity

    Whether the stored document is plausibly the kind of file the artifact declares, rather than something fetched by mistake.

  • passType matchnot applicable to this artifact type

    Whether the artifact is really the kind of thing its metadata claims it is.

  • passFreshness

    How long since the source repository was last pushed to.

  • passPrompt injection

    Scans the artifact's own text for instructions aimed at your agent rather than at you.

  • passLicense

    Whether the source repository declares an SPDX license permissive enough to redistribute.

How the grade is calculated

Each check contributes 0 points when it passes, 1 when it warns, and 2 when it fails. The total maps to a letter:

  • Aevery check passed
  • Bone warning
  • Ctwo warnings
  • Dprompt injection or body integrity failed, or three warnings
  • Fone of those failed, and something else is wrong

These are automated hygiene checks, not a security audit, and not a dependency or vulnerability scan. A grade of A means nothing was flagged — not that the artifact is safe.

Versions

  • git-e94b47e822b52026-07-31