Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Ai Executor Zero Job Contract

ASecurity

讀或修改 starledger canonical prompt 的 step 3、診斷任何 zero-job run 的報告文字、或有人(含 executor 本身)提議「backlog 排空了/停用 routine/P3 完成」時載入。這是 zero-job 結論的權威邊界與逐字報告契約。

2 stars
0 votes
0 copies
3 views
Added 9/4/2026
ai-agentsgobashnodegit

Security Analysis

A100/100

Scanned 9/4/2026

$npx -y skills add F-e-u-e-r/opus-pack --skill ai-executor-zero-job-contract --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ai Executor Zero Job Contract?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Ai Executor Zero Job Contract
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/f-e-u-e-r-ai-executor-zero-job-contract/badge)](https://www.skillsdirectory.com/skills/f-e-u-e-r-ai-executor-zero-job-contract)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: ai-executor-zero-job-contract
description: 讀或修改 starledger canonical prompt 的 step 3、診斷任何 zero-job run 的報告文字、或有人(含 executor 本身)提議「backlog 排空了/停用 routine/P3 完成」時載入。這是 zero-job 結論的權威邊界與逐字報告契約。
---

# starledger:executor zero-job 契約(3a–3e)

(驗證日 2026-07-30,starledger HEAD `55ae7b2`。權威文本=
`docs/routines/starledger-ai-classification.md` canonical prompt 的 step 3;本 skill 是導讀與
邊界,不是替代品——兩者衝突時以 repo 文本為準。)

## 事故背景(為什麼這份契約長這樣)

2026-07-26:一次 zero-job run 把「當下正確的 668/668 快照」升格為常設宣稱——「fully
drained」「disable this routine」「P3 完成閘達成」——34 分鐘後 daily sync +17 顆星就全數否證
(PR #205 修正;starledger git 歷史可重建時間軸)。兩個教訓都寫進了契約:

1. **快照結論不得升格為常設結論**。star 語單持續成長,「排空」本質上是暫時狀態。
2. **planner 是 budget-limited 的**(`config/ai.yaml` 的 `max_retry_per_run`/
   `max_refresh_per_run` 現值 0):zero-job manifest 在邏輯上**無法**證明「所有 annotation 都在
   當前 fingerprint」——refresh/retry 候選被 ceiling 靜默丟棄。所以合法的最強宣稱只有
   「snapshot-local manifest is empty **under this base and config**」。

## 協定地圖(依序、遇第一個失敗即停)

前置條件:step-2 `plan` 指令本身 exit 0 且本輪確實寫出 `.ai-runs/manifest.json`;否則報該失敗
指令、exit 0、no conclusion。

| 步 | 檢查 | 通過 | 失敗(全部 fail-closed、exit 0) |
|---|---|---|---|
| 3a | `git fetch --no-tags origin main` 後 `git diff --quiet "$BASE_SHA" origin/main -- stars.json dataset-meta.json ai-annotations.json ai-annotations-meta.json`(`BASE_SHA`=run 開始時記錄的 commit;step 11 前 HEAD 不得移動,故可 `git rev-parse HEAD` 重導出) | diff exit 0 → 3b | fetch 失敗→`zero jobs; base freshness unverifiable (fetch failed); no conclusion`;diff exit 1→逐字 `empty manifest on stale base; base advanced during run; re-run` + 兩個 SHA;其他 exit→`… (git diff failed) …` + exit 碼 |
| 3b | `config/ai.yaml`:結構檢查**先**(檔可讀+`ai.enabled`/`ai.executor_kind`/四個 budget ceiling——`max_new_per_run`/`max_retry_per_run`/`max_refresh_per_run`/`max_total_per_run`——齊且型別對)、值檢查**後**(enabled true、kind `claude-routine`、`max_total_per_run`>0) | → 3c | 結構壞→`zero jobs; config unverifiable; no conclusion`;值壞→報**具體**原因,不得說 backlog/manifest empty |
| 3c | node_id **集合**相等(count 相等不是證據)+ omitted-unfetchable=0(只認**完整**的 step-2 stdout capture;截斷/不可解析=unavailable) | 全零 → 3d | 任一非零→報五個數字、永不說 "fully classified";stdout 不可得→`zero jobs; omitted-unfetchable status unavailable; no conclusion`,**絕不原地重跑 plan**(第二份 live 觀察會覆寫 manifest) |
| 3d | (3a–3c 全過)唯一允許的正向結論 | 逐字 `snapshot-local manifest is empty under this base and config` + 證據塊:BASE_SHA、manifest `dataset_sha256`、五數字、四個 budget ceiling、omitted(0)、jobs(0)。cadence 放寬註記為 **3d 專屬** | — |
| 3e | 3c 的**唯一例外**:missing=0 ∧ duplicates=0 ∧ extra>0 ∧ omitted=0 ∧ jobs=0 → removed-star prune 維護路徑 | 見 `removed-star-prune-lifecycle` skill | 收據不完整/數字不合→fail closed + revert |

## 權威邊界(每個 zero-job 結局都適用)

- executor **永不**建議停用本 routine(最多:3d 情況下註記 operator *可*考慮放寬 cadence)。
- executor 對**任何** milestone 都沒有完成權威;P3 的完成**專屬** operator 手動 dispatch 的
  read-only `p3-completion-check` workflow(見 `p3-completion-and-closeout` skill)。
- 逐字報告字串是**介面**:executor prompt、`scripts/p3-completion-check.mjs` 的
  regex(`/omitted (\d+) probe-ok/`)與 operator 通知都在解析它們。改任何一句=介面變更,
  必須 sweep 全部消費者(operational-rigor 的 call-site sweep)。

## 修改 step 3 的正確程序

1. 改 `docs/routines/starledger-ai-classification.md` 的 fenced prompt(機械稽核見
   `verification-battery-and-doc-audits` skill)。
2. 走正常 PR + CI + owner 的 review 陣容(見 `review-and-merge-workflow` skill)。
3. **合併後 live prompt 必須 reconcile**——repo 內的 prompt 只是 canonical 文本,live routine
   持有副本;不 reconcile 修正就是 inert(見 `live-routine-reconciliation` skill)。

**完成定義**:三步全做完+下一輪 scheduled run 的報告文字符合新契約(owner 的 push 通知是
run 報告的觀察通道;repo 端 zero-job run 不留痕跡)。

## 負例(本會話實際發生過的不安全合理化,勿重演)

- ❌「668 annotations == 668 repositories,backlog fully classified——叫 operator 停用
  routine。」(事故原文的等價敘述;count 相等不是集合相等,快照不是常設狀態,executor 沒有
  停用建議權。)
- ❌「這次 full-corpus plan 0 jobs 就是 P3.5 在等的完成閘。」(完成閘是 operator dispatch 的
  workflow+記錄程序,executor 的 0 jobs 最多是 snapshot-local 證據。)
- ❌「stdout 不見了,重跑一次 plan 讀 omitted 就好。」(owner 駁回:第二次 plan 是另一份 live
  觀察且會覆寫 manifest——fail closed 報 no conclusion。)

## 再驗證

```bash
# 逐字字串逐條驗(合併式 alternation 會遮蔽單一字串消失;逐條才看得到缺哪個)
for s in \
  "snapshot-local manifest is empty under this base and config" \
  "empty manifest on stale base; base advanced during run; re-run" \
  "Never skip 3a" \
  "omitted-unfetchable status unavailable"; do
  rg -q "$s" docs/routines/starledger-ai-classification.md || echo "MISSING: $s"
done   # 無輸出 = 契約在場;任何 MISSING = 契約被動過,先讀 diff 再行動
```

Attribution

F-e-u-e-rF-e-u-e-r
View sourceSee grades on GitHubMore from F-e-u-e-r →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698621 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →