# Verify Workflow

> Use when the human wants to validate verification evidence and update the checklist with final status. Reads the testcase definitions and verification record, validates completeness, and updates the checklist. Do not use to write production code, run tests, or repair failures.

- Skill: `phananhtuan09/verify-workflow` (Agent Skill, multi-file: 9 files)
- Install (CLI): `npx skillmds@latest add phananhtuan09/verify-workflow`
- Raw SKILL.md: https://api.skillmd.com/api/skills/phananhtuan09/verify-workflow/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: phananhtuan09 (https://skillmd.com/u/phananhtuan09)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/phananhtuan09/verify-workflow

---


# Verify Workflow

Validate verification evidence and update the checklist with final status.

This skill is the final step in the verification pipeline.
It does not run tests, drive browsers, or collect evidence itself.
It reads the artifacts produced by `verify-feature` and `verify-runtime`, validates their completeness, and updates the checklist.

## Flow

```
1. manual-checklist  → creates checklist.md + testcase-definitions.json
2. verify-feature    → appends implementation evidence to verification-record.md
3. verify-runtime    → appends E2E evidence to verification-record.md
4. verify-workflow   → validates → updates checklist.md
```

This skill runs step 4.

## Input

- Required: feature slug, for example `{feature-name}`.
- Required: testcase definitions path, for example `docs/ai/features/checklists/{feature-name}-testcases.json`.
- Required: verification record path, for example `docs/ai/features/verifications/{feature-name}.md`.
- Required: checklist path, for example `docs/ai/features/checklists/{feature-name}.md`.

## Output

Update `docs/ai/features/checklists/{feature-name}.md` with final evidence status.

## Validation Phase

Before updating the checklist, validate the artifacts:

### 1. Validate testcase-definitions.json schema

Check that the JSON file contains:
- `feature` field (string)
- `spec_path` field (string)
- `testcases` array (non-empty)
- Each testcase has: `id`, `ac`, `test_type`, `steps`, `expected`, `done_criteria`
- Each `done_criteria` has: `required` (array), `not_sufficient` (array)
- `test_type` values are defined in the project config

If validation fails, stop and report:
```
Planning defect: testcase-definitions.json has invalid schema.
{specific error}
Fix the testcase definitions before running verification.
```

### 2. Validate verification-record.md completeness

For each testcase in the JSON, check that the verification record contains evidence:
- Search for the testcase ID (e.g., `TC-001`) in the verification record
- If not found, the testcase has no evidence
- If found, check that evidence matches `done_criteria.required`

### 3. Cross-check test_type assignment

- `code_test`, `build_check`, `api_check` testcases should have evidence from `verify-feature`
- `runtime_e2e` testcases should have evidence from `verify-runtime`
- If a testcase type has no matching evidence section, record it as `🔴 Chưa chạy`

## Checklist Update Phase

Read `skills/verify-workflow/references/evidence-rules.md` for the complete evidence rules.

For each testcase in the JSON:

1. Find matching evidence in the verification record
2. Classify evidence status using the rules from evidence-rules.md:
   - `🟢`: All `done_criteria.required` items satisfied, no `not_sufficient` items matched
   - `🟡`: Some evidence exists but incomplete or indirect
   - `🔴`: No evidence, or evidence contradicts expected result
3. Update the checklist line:
   - Change icon from `🔴` to `🟢` or `🟡`
   - Update AI verification note to one short method-and-result phrase
   - Remove `[ ]` task marker only for green testcases

For regression testcases (`RG-*`):
- Apply the same evidence classification rules
- Keep regression counts separate from spec testcase counts

## Summary Update

After updating all testcase statuses, update the checklist summary section:

```markdown
## Tóm tắt xác minh
- Tổng số ca kiểm thử: {N}
- 🟢 Đã xác minh đầy đủ: {green}/{N} ({green_percent}%)
- 🟡 Có bằng chứng một phần: {yellow}/{N} ({yellow_percent}%)
- 🔴 Cần người kiểm tra: {red}/{N} ({red_percent}%)
- Tiêu chí chấp nhận được bao phủ đầy đủ: {covered_ac}/{total_ac}
- Bằng chứng chi tiết: docs/ai/features/verifications/{feature-name}.md
```

Update the `## Rủi ro chưa kiểm` section if there are untested surfaces.

## Final Status

Determine the overall verification status:

- `ĐẠT`: all testcases are green
- `KHÔNG ĐẠT`: at least one testcase is red with confirmed failure
- `CẦN BẠN XÁC NHẬN`: at least one testcase is yellow
- `BỊ CHẶN`: verification could not run due to missing environment or artifacts

## Human-Facing Output

Return a short status in Vietnamese:

```text
KẾT QUẢ: {ĐẠT | KHÔNG ĐẠT | CẦN BẠN XÁC NHẬN | BỊ CHẶN} — {lý do ngắn}

🟢 {n}/{total} spec   🟡 {n}/{total} spec   🔴 {n}/{total} spec
🟢 {n}/{total} regression

  🔴 {TC-ID}  {một dòng: thực tế sai gì}
  🟡 {TC-ID}  {một dòng: thiếu bằng chứng gì}

Checklist: docs/ai/features/checklists/{feature-name}.md
```

Rules:

- List only non-green testcases, one line each.
- Omit the regression line when the run produced no regression testcase.
- When nothing is red or yellow, return at most three lines plus the checklist path.
- Never claim a verification the validation marked insufficient.
- When blocked, say what was unavailable and state that no checklist status was changed.

## Handoff

Offer `handoff` on the next line only when at least one testcase is red with confirmed failure.

The handoff prompt should reference:
- The specific testcase ID
- The evidence that shows the failure
- The expected behavior from the testcase definition

Route by failure type:
- Implementation defect → target `/fix-bug`
- Missing scope from spec → target `/execute-spec`
- Unclear expected result → state which expected result must be decided first

## Artifact Boundaries

- Do not modify code, tests, specs, or testcase definitions.
- Do not run tests or drive browsers.
- Do not repair failures.
- Only update the checklist file based on evidence in the verification record.

## Orchestrator Contract

When this skill is run under `/orchestrator`, append exactly one HTML comment as the final output line:

- No confirmed failure:
  `<!-- orchestrator: outcome=continue provides=checklist_path checklist_path=docs/ai/features/checklists/{feature-name}.md -->`
- At least one confirmed failure:
  `<!-- orchestrator: outcome=stop-fail -->`
- Required inputs or environment prevented validation:
  `<!-- orchestrator: outcome=stop-blocked -->`

Rules:

- Emit the comment only after the checklist has been updated.
- `checklist_path` must match the file actually updated.
- If this skill runs standalone, the comment is optional.