Claude-Code-Game-Studios/CCGS Skill Testing Framework/skills/utility/smoke-check.md

Skill Test Spec: /smoke-check

Skill Summary

/smoke-check runs the critical path smoke test checklist for a build. It reads the QA plan from production/qa/ and checks each critical path item against the acceptance criteria defined in the current sprint's stories. Items that can be evaluated analytically are assessed; items that require runtime verification or visual inspection are flagged as NEEDS MANUAL CHECK.

The skill produces no file writes — output is conversational. No director gates apply. Verdicts: PASS (all critical items verified), FAIL (at least one critical item fails), or NEEDS MANUAL CHECK (critical items exist that require human verification).

---

Static Assertions (Structural)

Verified automatically by /skill-test static — no fixture needed.

• Has required frontmatter fields: name, description, argument-hint, user-invocable, allowed-tools
• Has ≥2 phase headings
• Contains verdict keywords: PASS, FAIL, NEEDS MANUAL CHECK
• Does NOT contain "May I write" language (skill is read-only)
• Has a next-step handoff (e.g., /bug-report on FAIL, /release-checklist on PASS)

---

Director Gate Checks

None. /smoke-check is a QA utility skill. No director gates apply.

---

Test Cases

Case 1: Happy Path — All critical path items verifiable, PASS

Fixture:

• production/qa/qa-plan-sprint-005.md exists with 4 critical path items
• All 4 items are logic or integration type (analytically assessable)
• Corresponding story ACs are defined and met per sprint stories

Input: /smoke-check

Expected behavior:

1. Skill reads the QA plan and identifies 4 critical path items
2. Skill evaluates each item against the story's acceptance criteria
3. All 4 items pass
4. Skill outputs a checklist: each item with a PASS marker
5. Verdict is PASS with summary: "4/4 critical path items verified"

Assertions:

• All 4 items appear in the checklist output
• Each item is marked PASS
• Verdict is PASS
• No files are written

---

Case 2: Failure Path — One critical item fails, FAIL verdict

Fixture:

• QA plan has 3 critical path items
• Item 2 ("Player health does not go below 0") fails — story AC indicates clamping logic was not implemented

Input: /smoke-check

Expected behavior:

1. Skill evaluates all 3 items
2. Item 1 and Item 3 pass; Item 2 fails
3. Skill outputs checklist with specific failure: "Item 2 FAIL — Health clamping not verified"
4. Verdict is FAIL
5. Skill suggests running /bug-report for the failing item

Assertions:

• Verdict is FAIL (not PARTIAL or NEEDS MANUAL CHECK)
• Failing item is identified by name/description
• Passing items are also shown (not hidden)
• /bug-report is suggested for the failure

---

Case 3: Visual Item Cannot Be Auto-Verified — NEEDS MANUAL CHECK

Fixture:

• QA plan has 3 items: 2 logic items (PASS) and 1 visual item ("Explosion VFX triggers correctly on enemy death" — ADVISORY, visual type)

Input: /smoke-check

Expected behavior:

1. Skill evaluates the 2 logic items — both pass
2. Skill evaluates the visual item — cannot be verified analytically
3. Visual item is marked NEEDS MANUAL CHECK with a note: "Visual quality requires human verification — see production/qa/evidence/"
4. Verdict is NEEDS MANUAL CHECK (not PASS, because human action is required)
5. Guidance on how to perform manual check is provided

Assertions:

• Verdict is NEEDS MANUAL CHECK (not PASS or FAIL)
• Visual item is marked with explicit NEEDS MANUAL CHECK tag
• Guidance for manual verification process is included
• Logic items are still shown as PASS

---

Case 4: No Smoke Test Plan — Guidance to run /qa-plan

Fixture:

• production/qa/ directory exists but contains no QA plan file for the current sprint
• Current sprint is sprint-006

Input: /smoke-check

Expected behavior:

1. Skill looks for QA plan for the current sprint — not found
2. Skill outputs: "No smoke test plan found for sprint-006"
3. Skill suggests running /qa-plan sprint-006 first
4. No checklist is produced

Assertions:

• Error message names the missing sprint's plan
• /qa-plan is suggested with the correct sprint argument
• Skill does not produce a checklist with no plan
• Verdict is not PASS (error state, no checklist evaluated)

---

Case 5: Director Gate Check — No gate; smoke-check is a QA utility

Fixture:

• Valid QA plan with assessable items

Input: /smoke-check

Expected behavior:

1. Skill runs the smoke check and produces a verdict
2. No director agents are spawned
3. No gate IDs appear in output

Assertions:

• No director gate is invoked
• No write tool is called
• Verdict is PASS, FAIL, or NEEDS MANUAL CHECK — no gate verdict involved

---

Protocol Compliance

• Reads QA plan before evaluating any items
• Evaluates each item explicitly (no silent skips)
• Visual/feel items are always flagged NEEDS MANUAL CHECK (not auto-passed)
• FAIL verdict triggers on first critical failure (not advisory)
• Verdict is PASS, FAIL, or NEEDS MANUAL CHECK — no other verdicts

---

Coverage Notes

• The case where the QA plan exists but has no critical path items (all items are ADVISORY) is not tested; PASS would be returned with a note that no critical items were checked.
• The distinction between BLOCKING and ADVISORY gate levels from coding-standards.md is relied upon to determine which items can produce a FAIL.
• Build-specific failures (runtime crashes) that occur during manual testing are outside the scope of this skill — use /bug-report for those.