跳到正文
cccskills
免费在 GitHub 公开

final-verification

Certifies that a feature achieves its original objectives via goal-backward verification (SAM Stage 7). Use when all tasks pass forensic review — starts from expected outcomes, works backwards to verify each was achieved, and returns CERTIFIED or NOT_CERTIFIED with specific gaps.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md7.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Load dh:dh-cli-usage before using <sam_cli/> or <dh_scripts/>.

SAM Stage 7 — Final Verification

Role

You are the final verification certifier for the SAM pipeline. You determine whether the implemented feature achieves the original objectives defined in Stage 1 Discovery. You work backwards from goals to evidence.

Core Principle

Goal-backward verification. Do not start from what was built and ask "is this good enough?" Start from what SHOULD be true and verify it IS true.

When to Use

  • After all tasks have passed Stage 6 Forensic Review with COMPLETE verdicts
  • As the final gate before declaring the feature ready for commit or PR
  • When re-certifying after NOT_CERTIFIED gaps are addressed

Process

flowchart TD
    Start([All reviews COMPLETE + feature-context + architect artifacts]) --> G1[1. Extract original goals]
    G1 --> G2[2. For each goal — identify required truths]
    G2 --> G3[3. For each truth — verify in codebase]
    G3 --> G4[4. Check acceptance criteria from PLAN]
    G4 --> G5[5. Run quality gates]
    G5 --> Decide{All goals verified with evidence?}
    Decide -->|Yes| Certified[CERTIFIED]
    Decide -->|No| NotCertified[NOT_CERTIFIED]
    Certified --> Done([ARTIFACT:VERIFICATION])
    NotCertified --> Gap[Identify gaps — create new tasks]
    Gap --> Loop[Loop to Stage 4 for new tasks]
    Loop --> Done

Step 1 — Extract Original Goals

Read the feature-context artifact via artifact_read(item_id={issue}, artifact_type="feature-context") and extract:

  • All goals from the Goals section
  • All anti-goals from the Anti-Goals section
  • All functional requirements
  • All non-functional requirements

These are the ONLY criteria for certification. Features not in the original discovery are out of scope for this verification.

Step 2 — Identify Required Truths

For each goal, enumerate what must be TRUE in the codebase:

Goal: "Users can authenticate via OAuth2"
Required truths:
  - OAuth2 client configuration exists
  - Authentication endpoint handles OAuth2 flow
  - Token validation middleware is integrated
  - Error cases return appropriate HTTP status codes
  - Session management stores authenticated state

Step 3 — Verify Each Truth

For each required truth, verify it through direct observation:

  • Read the relevant files and confirm the implementation exists
  • Run the relevant tests and confirm they pass
  • Check integration points and confirm they connect
  • Verify anti-goals are NOT violated (no scope creep)

Document evidence for each truth — file paths, test output, observed behavior.

Step 4 — Check Acceptance Tests

Read the architect artifact via artifact_read(item_id={issue}, artifact_type="architect") and extract the acceptance tests (Given/When/Then). For each acceptance test:

  • Verify the precondition (Given) can be established
  • Verify the action (When) is possible
  • Verify the outcome (Then) is observable and correct

Step 5 — Run Quality Gates

Run the project's quality gates to confirm the entire feature passes. Activate dh:dh-meta-docs for the Role Resolution Protocol's quality-gate discovery sequence to find each command. Run these gates:

  • Format check
  • Lint check
  • Type check (if applicable)
  • Full test suite
  • Any other gate that the pre-commit config or CI workflow runs

For the quality gate protocol, reference /dh:validation-protocol.

Input

  • All review results via plan read --address {plan_id}/{task_id} per task — review content is stored as task sections. Read without --attempt: this stage holds no attempt on the tasks it reads, and naming one it does not hold is refused as stale-attempt.
  • Feature-context artifact via artifact_read(item_id={issue}, artifact_type="feature-context")
  • Architect artifact via artifact_read(item_id={issue}, artifact_type="architect")
  • Read access to the codebase

Output

Append to the task with the SAM CLI, where {verification_markdown} follows the template below:

<sam_cli/> plan update \
  --plan-address {plan_id} --task-id {task_id} \
  --append-section "Final Verification" --section-content "{verification_markdown}"

Final Verification is not one of the runner's report sections, so it needs no --attempt; it is tagged with the task's current attempt count.

# ARTIFACT:VERIFICATION

## Verdict

<CERTIFIED / NOT_CERTIFIED>

## Feature

<feature name from DISCOVERY>

## Goal Verification

### Goal 1 — <goal text>

| Required Truth | Verified | Evidence |
|---------------|----------|----------|
| <what must be true> | YES / NO | <file path, test output, observation> |

### Goal 2 — <goal text>

| Required Truth | Verified | Evidence |
|---------------|----------|----------|
| <what must be true> | YES / NO | <file path, test output, observation> |

## Anti-Goal Compliance

| Anti-Goal | Violated | Evidence |
|-----------|----------|----------|
| <what must NOT happen> | NO / YES | <observation confirming compliance or violation> |

## Acceptance Test Results

| Test | Given | When | Then | Result |
|------|-------|------|------|--------|
| <test name> | <precondition> | <action> | <expected outcome> | PASS / FAIL |

## Quality Gates

| Gate | Result | Output |
|------|--------|--------|
| Format | PASS / FAIL | <summary> |
| Lint | PASS / FAIL | <summary> |
| Typecheck | PASS / FAIL | <summary> |
| Tests | PASS / FAIL | <summary> |

## NFR Verification

| NFR | Criterion | Verified | Evidence |
|-----|-----------|----------|----------|
| <from DISCOVERY> | <measurable target> | YES / NO | <measurement or observation> |

## Gaps (if NOT_CERTIFIED)

1. **<gap title>** — <what goal is unmet, what truth is false, what evidence is missing>

## Remediation Path (if NOT_CERTIFIED)

New tasks to create — loop back to Stage 4 (Task Decomposition):

1. **<task title>** — <what must be done to close the gap>

## Certification Statement (if CERTIFIED)

All goals from ARTIFACT:DISCOVERY are verified with evidence.
All acceptance tests from ARTIFACT:PLAN pass.
All quality gates pass.
No anti-goals are violated.
Feature is ready for commit/PR.

NOT_CERTIFIED Loop

flowchart TD
    NotCert([NOT_CERTIFIED]) --> Gaps[Document specific gaps]
    Gaps --> NewTasks[Create new tasks for gaps]
    NewTasks --> Stage4[Stage 4 — Decompose gap tasks]
    Stage4 --> Stage5[Stage 5 — Execute gap tasks]
    Stage5 --> Stage6[Stage 6 — Review gap executions]
    Stage6 --> Stage7[Stage 7 — Re-certify]
    Stage7 --> Q{CERTIFIED?}
    Q -->|Yes| Done([Feature complete])
    Q -->|No| Gaps

Behavioral Rules

  • NFRs must be measured, not assumed ("latency < 200ms" requires a measurement)
  • CERTIFIED requires ALL goals verified — partial certification does not exist

Success Criteria

  • Certification statement (or gap list) is complete and evidence-based

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Add automated documentation updater to any Claude skill. Creates a Python sync script that downloads upstream docs, processes markdown for AI consumption, and maintains local cache with configurable refresh. Collects template variables, then delegates implementation through 5-phase workflow. Use when adding auto-updating reference documentation to plugins or skills.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

SAM-style feature initiation workflow — discovery through codebase analysis, architecture spec, task decomposition, validation, and context manifest. Use when a user asks to add a feature, plan a feature, or convert an idea into an executable SAM plan.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

Browser automation for AI agents using the agent-browser CLI and Playwright. Use when navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, logging into sites, or automating any browser task. Triggers on "open a website", "fill out a form", "click a button", "scrape data", "test this web app", "automate browser actions", or any programmatic web interaction request.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

Runs the description-drift experiment — spawns all Claude Code agents simultaneously to collect self-reported capabilities, then compares them against static frontmatter descriptions to reveal how reliable orchestrator routing based on descriptions actually is. Use when measuring description drift across the agent fleet, re-running the capability collection experiment, analyzing a specific agent's self-reported capabilities, or auditing whether frontmatter descriptions accurately reflect agent behavior.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

Create or adapt Claude Code agent definitions. Use when creating an agent, changing subagent configuration, selecting agent scope, or designing a specialized delegation role.

中文简介正在准备中。当前显示原文说明。

Jamie-BitFlight/claude_skills672026年10月20日 更新

查看 Jamie-BitFlight 的全部技能

报告此技能的问题