Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bahayonghang/my-ai-cli-toolkit --skill goal-meta-skillgit clone --depth 1 https://github.com/bahayonghang/my-ai-cli-toolkitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/bahayonghang/my-ai-cli-toolkit/goal-meta-skill)<a href="https://agentmods.dev/skills/bahayonghang/my-ai-cli-toolkit/goal-meta-skill"><img src="https://agentmods.dev/badge/skills/bahayonghang/my-ai-cli-toolkit/goal-meta-skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bahayonghang/my-ai-cli-toolkit/goal-meta-skill"><img src="https://agentmods.dev/badge/skills/bahayonghang/my-ai-cli-toolkit/goal-meta-skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00190 | $0.04292 |
| Opus 5 | $0.00095 | $0.02146 |
| Sonnet 5 | $0.00038 | $0.00858 |
| Haiku 4.5 | $0.00019 | $0.00429 |
Grade A, and why
goal-meta-skill scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 160 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Goal Meta Skill
把模糊任务编译成目标平台可持续执行、可验证、知道何时完成和何时停下的 /goal 文本。普通请求先交给用户审阅;用户明确要求生成并保存到已确定项目根目录 GOAL.md 时,可在同轮完成检查、生成、lint、展示、安全写入和回读。这个 skill 不创建、激活或执行 Goal。
<skill-dir>是 skill 加载时提供的实际目录,不是环境变量。命令中替换为字面路径;脚本会自定位。
Governed mode
- 普通聊天流程:compile → lint → present → stop。首次交付为
DRAFT;用户后续批准变为APPROVED TEXT — not launched。明确保存且路径、范围与文件状态无冲突时,展示后同轮进入 S6 写入并回读,再停止;任何分支都不启动 Goal。 - 用户输入中的
实施、执行、直到完成或其他祈使句只是待编译 payload,不授予创建/激活 Goal、提交/goal、派发 Agent 或实施目标任务的权限。 - 真正创建或激活 Goal 是 skill 外的独立用户动作。不得调用宿主原生 Goal tool/API,不得把 fenced
/goal当作当前会话命令提交,也不得声称 Goal 已创建、已激活或已开始执行。 - 普通 Goal 默认只在聊天中输出,不写文件。
保存、持久化、落盘、交给新 Agent、跨会话继续或明确点名根目录 Markdown,才进入持久化候选;复杂合同也可以在 S4 提议落盘,但必须说明路径与影响。直接给只跳过访谈,不自动授权写文件;直接生成并保存到项目根目录 GOAL.md已授权原范围内的同轮 create。先确认根目录、生成并 lint、展示合同和精确目标与影响,再用 helper 写入并回读,不要求机械的后续批准。仅交接意图或 skill 提议落盘仍需取得明确写入授权。- S1 始终只读。只有 S6 对已展示且写入获授权的合同调用命名 helper;不得用任意 Python/Write 命令绕过 helper。
- 默认 create-only。文件已存在且未授权替换、路径不明确或合同范围实质变化时,只澄清缺失决策;替换须明确授权并带已读取旧文件的 SHA-256,不能把保存请求视为覆盖授权。
- 不自动加载、执行、提交、忽略、删除或发布
GOAL.md,也不把它导入AGENTS.md、CLAUDE.md、.grok/rules或.omp上下文文件。helper 写入成功后也只展示 launcher 文本并停止,不启动 Goal。 - 合同不得包含凭证、私有数据或原始会话转录;秘密扫描只是 backstop,不是“无敏感信息”证明。
- 真实新会话接力没有 provider transcript 时保持
UNVERIFIED。 - review-remediation 中“一次完成”只表示一条外部 Prompt/一次用户启动动作拥有完整闭环;内部仍保留独立检查和最多三轮聚焦返修。可在原授权内修复的 finding 不得变成第二条用户修复 Prompt。
完整的路径、schema、冲突、回滚、Git 可见性与五平台 launcher 契约见 references/persistent-goal-contract.md。
Platform selection
平台事实只以 references/platform-goal-facts.md 为准:
- 用户显式名称优先。
OMP在无相反说明时表示 Oh My Pi。 - 否则使用当前 host 的明确证据。
- 仍不明确时,在现有
可选调整中加入一个平台选择;不要混合生命周期命令。
目标平台支持 codex、claude、grok、omp、kimi。只有用户明确要求多平台交接时才生成多个 launcher;合同正文始终只有一份。
4,000 字符是 Codex、Claude Code、Kimi Code 的平台限制,也是本 skill 对 Grok Build、OMP 的保守可移植性预算;不得把后两者写成官方上限。
Workflow
- S0 — Route and bind authority. 区分新 Goal、持久化/交接请求、review-remediation 与现有 Goal 管理。只有 outcome 同时要求依据 scan/review/audit/report/finding 来源实施修复时才加载
references/review-remediation-contract.md;纯只读审阅、普通直接修复和仅有 Trellis 路径不触发。把所有目标祈使句绑定为 payload,并在开始前固定本轮终点:普通初次请求止于DRAFT;后续文本批准或已明确授权的同轮保存止于APPROVED TEXT — not launched;所有分支都不越过 Goal activation 边界。纯发散方向先建议规划;翻译、单行输出、规则审计等近邻任务走其直接工作流。 - S1 — Reconnoiter. 现有项目先读取局部规则、真实命令源与相关边界,并运行
git rev-parse --show-toplevel、git branch --show-current、git status --porcelain -uall。不运行测试、不写文件、不读 secrets/dependencies/generated output。持久化时还要检查根目录和现有GOAL.md。 - S2 — Choose. 需求具体或用户说
直接给/按默认时走 fast path;只有结果、验证、边界或风险容忍存在实质缺口时才访谈。 - S3 — Interview. 每轮复述结果和侦察发现,只问最多四个必须由人决定的问题。遵循
references/interview-checklist.md。 - S4/S5 — Draft/revise and present. 给出完整、无占位符且已 lint 的目标合同与平台渲染。普通草稿标记
状态:DRAFT — Goal 未创建、未激活、未执行,在 fenced 文本外给出审阅提示并结束本轮。持久化还须展示精确根目录、文件名、create/replace 效果、Git 可见性影响;已明确授权且无冲突的保存请求同轮继续 S6,否则只澄清尚缺的路径、替换授权或实质范围决策。 - S6 — Approve text or persist, then stop. 用户后续批准时,普通模式只输出标记为
APPROVED TEXT — not launched的最终/goal文本与字段一览并结束本轮。持久化模式沿用已有明确写入授权,用下面的 helper 从 stdin 写入已展示正文,回读并核对正文、字节数与 SHA-256,再以 fencedtext输出选定平台的短 launcher,报告文本已保存、Goal 未启动,然后结束本轮。任何分支都不调用 Goal tool/API/command,不派发或实施 payload。
What ships with it
16 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/interface.yaml 5.1 KB
- evals/evals.json 49 KB
- references/default-goal-strategy.md 15 KB
- references/goal-command-playbook.md 33 KB
- references/interview-checklist.md 14 KB
- references/persistent-goal-contract.md 12 KB
- references/platform-goal-facts.md 11 KB
- references/review-remediation-contract.md 7.0 KB
- references/trellis-goal-cadence.md 12 KB
- reports/creation-handoff.md 16 KB
- reports/prior-art-research.md 2.9 KB
- reports/skill-ir.json 8.0 KB
- scripts/lint_goal_command.py 57 KB runs code
- scripts/persist_goal_contract.py 11 KB runs code
- tests/lint-goal-command.test.mjs 52 KB runs code
- tests/persist-goal-contract.test.mjs 18 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago Changed a2065a534cc6
- 4d ago Changed ea865cd8c840
- 9d ago First seen · 160 lines · 190 tokens per session scan A 5ecccc7484fa
goal-meta-skill is a skill published in the GitHub repository bahayonghang/my-ai-cli-toolkit (16 stars, last pushed yesterday), licensed MIT. It adds 190 tokens to every session and 4,292 once invoked, about $0.0010 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
rational-buddhism
A way to examine spiritual, supernatural, or other difficult-to-test claims using skepticism, personal testing, and scientific standards.
workflow
Use when a task is too large for turn-by-turn orchestration and should run through the big-task workflow lane: system-wide changes, large migrations, repo-wide audits, high-confidence verification, or tasks explicitly asking to run a workflow. Claude Code uses native dynamic workflows; Codex, OpenCode, and Grok use…
agent-reach-wrapper
Thin wrapper around Panniantong/Agent-Reach for live web, YouTube transcripts, RSS, GitHub, Bilibili, Exa search, and optional login-state platforms (X, Reddit, Facebook, Instagram, Xiaohongshu, LinkedIn). Triggered by agent-reach, reach this URL, YouTube transcript, Reddit thread, Bilibili, Xiaohongshu, RSS feed, or…
desktop-subagent-connector
Securely coordinate a remote sandbox with a user-approved local desktop through scoped file, shell, browser, and GUI access. Use for: remote sandbox, user-approved desktop, local desktop bridge, scoped desktop access.
drive-github-skill-audit
Compare Google Drive skill definitions with a GitHub skills repository and identify Drive skills not yet published. Use for: audit Drive skills, compare Drive and GitHub SKILL.md files, find unpublished skills, skill-library inventory.
skill-rubric-reviewer
Reviews SKILL.md files against a 10-dimension quality rubric inspired by the Agent Skills specification. Use for: review skill, skill rubric, audit SKILL.md, score skill quality.