AgentStack
SKILL verified MIT Self-run

Stimuli Unit Function Review

skill-cuteweather-agent-stimuli-unit-function-review · by cuteweather

run_unit_review 的两阶段人工审核门禁。第一次生成 gate 并暂停,人工填写后再 finalize 生成 unit_function_review.md。

No reviews yet
0 installs
4 views
0.0% view→install

Install

$ agentstack add skill-cuteweather-agent-stimuli-unit-function-review

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Stimuli Unit Function Review? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Unit Function Review(Manual Pause/Resume)

目标

run_unit_review 必须在同一个 action 内支持两阶段:

  1. request_manual_review:只生成/刷新 gate,并触发 workflow pause。
  2. finalize_manual_review:读取人工填写后的 gate,生成最终 unit_function_review.md,并给出 passed/fallback 结论。

禁止新增 action 名,禁止新增平行 step。

固定输入与输出

  • gate 路径固定:/script/unit_function_review_gate.json
  • 最终报告路径固定:/unit_function_review.md

阶段判定(只看 gate)

  • 若 gate 不存在,或存在但未标记人工完成:进入 request_manual_review
  • 若 gate 已标记人工完成:进入 finalize_manual_review

人工完成判定规则:

  • human_review_completed == true,且
  • review_status == "finalized"

gate 协议(必须包含)

{
  "review_status": "pending_manual_review | finalized",
  "human_review_completed": false,
  "passed": false,
  "fallback_needed": true,
  "summary": "string",
  "findings": [],
  "checks": {},
  "vlm_review": {},
  "critical_mismatches": [],
  "next_actions": []
}

可直接参考模板:references/unit_function_review_gate.template.json

说明:

  • gate 是唯一人工输入入口。
  • request 阶段可补充 skeleton 与运行元信息,但不得伪造“已人工完成”。
  • finalize 阶段不得覆盖人工已填写 verdict(尤其 passed/fallback_needed/summary/findings)。

Phase A: requestmanualreview

必须执行:

  1. 生成或刷新 script/unit_function_review_gate.json skeleton。
  2. 若 gate 已存在且包含人工填写字段,仅允许补充非冲突元数据,不覆盖人工 verdict。
  3. 不生成 unit_function_review.md

返回 JSON(必须):

{
  "status": "reviewing",
  "success": false,
  "summary": "unit manual review requested; fill gate then resume",
  "artifacts": {
    "unit_function_review_gate": "/script/unit_function_review_gate.json"
  },
  "evidence": {
    "files_read": [],
    "files_written": ["/script/unit_function_review_gate.json"],
    "commands_run": [],
    "subagents": []
  },
  "decision": {
    "retryable": false,
    "next_step": null,
    "terminate_workflow": false,
    "pause_workflow": true,
    "resume_step": "meta_orchestrator",
    "pause_reason": "manual_review_pending"
  },
  "error": null
}

Phase B: finalizemanualreview

必须执行:

  1. 读取人工 finalized gate。
  2. 基于 gate 内容生成 unit_function_review.md(报告内容必须来自人工 gate,不得重新自动 VLM 判定覆盖)。
  3. 根据 gate 的 passed/fallback_needed 生成最终 verdict。

finalize verdict 规则:

  • passed=truefallback_needed=false -> status=succeeded
  • 其他情况 -> status=failed_retryable

返回 JSON(必须):

  • artifacts 至少包含 gate 与 unit_function_review.md 路径。
  • 若失败,summary/error 需明确指出由人工 gate 判定触发 fallback。

明确禁止

  • 禁止在本技能内等待人工输入(轮询/阻塞)。
  • 禁止在 finalize 阶段重跑自动 VLM 并覆盖人工结论。
  • 禁止把 pause 伪装为 failedterminal/terminateworkflow。
  • 禁止更改 gate/report 固定路径。

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.