feat(skills): add public visual-qa skill for web UI review

[fabiolavillatoro2003]

Intent

The developer (acting as a firstmate crewmate on firstmate's own repo) tasked the agent with creating a new public, installer-facing skill at skills/visual-qa/SKILL.md that distills UI-review knowledge currently duplicated in two lower-fidelity places: an exhortation in the global CLAUDE.md and a concrete checklist in trashtalknyc-website's docs/agent-protocol.md. Requirements: base the work on an existing scout audit report (candidate #2) as rationale/content source; include the breakpoint matrix (1440/1280/768/375px), an interactive/hover/empty/success state checklist, the "passing tests does not imply visual quality" framing, a subjective-feedback translation table, and a DX-report closing format; keep it project-agnostic (not trashtalknyc-specific) and placed in public skills/ (not agent-only .agents/skills/), matching the skills/stow precedent, with no "internal" metadata and a clear trigger/description. The developer explicitly said not to touch the trashtalknyc-website repo/clone from this worktree — that doc-shrinking follow-up should just be noted for a separate task. Work had to follow firstmate-coding-guidelines (knowledge placement, shellcheck, one-sentence-per-line markdown, plain dash, no co-author), be committed on a dedicated branch, and pushed/validated through the no-mistakes pipeline; a GitHub push-permission blocker was resolved by the developer configuring a fork, after which they asked the agent to rerun the pipeline from the failed push step.

What Changed

• Add skills/visual-qa/SKILL.md, a new public, project-agnostic skill for reviewing web UI changes: a breakpoint matrix (1440/1280/768/375px), an interactive/hover/empty/success state checklist, a "passing tests does not imply visual quality" framing, a subjective-feedback translation table, and a structured closing report format.
• Update README.md to reference the new skill alongside existing public skills.

Risk Assessment

✅ Low: The change is documentation-only (a new public, installer-facing SKILL.md plus a two-line README update), closely mirrors the existing skills/stow precedent in frontmatter shape, style, and public/agent-only placement, stays project-agnostic as required, and does not touch any project clone or executable code.

Testing

The baseline shell test suite (47 tests/*.test.sh files covering bin/ scripts) already passed and is unaffected by this change, since the diff only touches README.md and a new public skill document. Since this is a documentation/content change with no runtime or rendered UI surface of its own, I verified the actual deliverable directly: the SKILL.md frontmatter parses correctly and matches the skills/stow public-skill precedent (no internal metadata, user-invocable), and the body contains every required element from the intent (breakpoint matrix, state checklist, "passing tests does not imply visual quality" framing, subjective-feedback translation table, closing report format), is project-agnostic, and correctly lives in public skills/ rather than .agents/skills/. Confirmed trashtalknyc-website was not touched and the working tree is clean. No issues found.

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

• command -v tmux >/dev/null || { echo "tmux is required for e2e tests" >&2; exit 1; }; tmux -V; rc=0; for t in tests/*.test.sh; do echo "== $t =="; bash "$t" || rc=1; done; exit "$rc"
• git diff 2061c37..e77af1c --stat — confirmed only README.md and skills/visual-qa/SKILL.md changed; projects/trashtalknyc-website untouched as required
• Parsed the SKILL.md YAML frontmatter — valid, matches the skills/stow public-skill precedent exactly (name, description, user-invocable: true, no metadata.internal)
• Grepped file content for required elements: breakpoint matrix (1440/1280/768/375px), the 'passing tests does not imply visual quality' framing, hover/interactive/empty/success state checklist items, the subjective-feedback translation table, and the 4-part closing report format — all present
• Confirmed the skill is project-agnostic (no trashtalknyc references) and correctly placed under public skills/ rather than agent-only .agents/skills/
• Checked README.md diff for consistency with the new skill and one-sentence-per-line markdown style
• bash tests/*.test.sh baseline suite (already run per task setup) — unrelated to this diff since no bin/ scripts changed
• git status --porcelain — working tree clean, no stray artifacts

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

[fabiolavillatoro2003]

Closing without merging - this is an internal-only change for our own firstmate installation, not an upstream contribution. Landing it directly on our local main instead.

[AdamTerhaerdt]

@coderabbitai review