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 agentmods add skills/loki-ai-ch/spec-manager/skillnpx skills add loki-ai-ch/spec-manager --skill skillgit clone --depth 1 https://github.com/loki-ai-ch/spec-managerWrote 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/loki-ai-ch/spec-manager/skill)<a href="https://agentmods.dev/skills/loki-ai-ch/spec-manager/skill"><img src="https://agentmods.dev/badge/skills/loki-ai-ch/spec-manager/skill.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00092 | $0.03377 |
| Opus 5 | $0.00046 | $0.01688 |
| Sonnet 5 | $0.00018 | $0.00675 |
| Haiku 4.5 | $0.00009 | $0.00338 |
Grade A, and why
spec-manager 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 5d 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
spec-manager — 本地项目迭代全链路
PRD → Design → Spec → Agent Task → 部署。纯本地，所有数据落 markdown + git。
用法
/spec-manager <需求>— 从需求开始走完整链路/spec-manager run <taskId>— 执行已有 Agent Task
新功能或非平凡工作先识别安全下一步，或生成 Agent Brief 收集本地上下文、历史 lessons、风险和下一步：
spec-manager next "<需求>"
spec-manager brief "<需求>"
安装 Agent 入口时优先推荐平台命令：spec-manager <platform> install，例如 spec-manager claude install、spec-manager codex install、spec-manager cursor install。跨框架安装使用 spec-manager agents install 或 spec-manager skills install。旧入口 spec-manager project agents --provider <provider> 保留给脚本和高级 provider 控制。
执行阶段可用 spec-manager assist next <taskId> --spec <L3-code> / spec-manager assist drift <taskId> --spec <L3-code> 对齐下一步和变更范围；验收前用 spec-manager assist acceptance <taskId> --spec <L3-code> 汇总证据、人工验收项和残余风险；最终回复前用 spec-manager assist delivery <taskId> --spec <L3-code> 生成面向用户的交付摘要。
UI、视觉或样式任务需要额外读取设计上下文：先运行 spec-manager brief "<需求>"，默认读取 resolved write root 的 specs/DESIGN.md，若不存在则兼容 fallback 到项目根目录 DESIGN.md。需求命中设计相关意图时，Agent Brief 会包含 Design Context 摘要、lint 概览和 source ref。涉及设计约束验收的 L3 可以声明 @verify: design-lint(specs/DESIGN.md) 或显式 @verify: design-lint(DESIGN.md)。看到 Design Context findings 时，先按 path 修复对应 DESIGN.md，再继续 UI 实现。注意：DESIGN.md 是产品视觉/设计上下文，不是 L2 Design 技术设计的替代品。
入口路由
收到 /spec-manager <用户输入> 时，按下表匹配子 skill；不确定时用当前工具的用户提问能力澄清。关键词重叠按 # 小者优先。
| # | 关键词 | 子 skill | 产出 |
|---|---|---|---|
| 1 | 修 typo / 改一行 / 改 log level | subskills/quick.md | quick |
| 2 | 查 / 看 / 列 / 统计 / 搜索 | subskills/research.md | (只读) |
| 3 | 需求 / 新功能 / 痛点 / 用户故事 | subskills/prd.md | L1 |
| 4 | 技术方案 / 架构 / 接口设计 | subskills/design.md | L2 |
| 5 | 实施 / 编码 / planJson | subskills/impl.md | L3 |
| 6 | 计划 / 里程碑 / 排期 | subskills/plan.md | plan |
| 7 | 测试方案 / QA / 测试用例 | subskills/testplan.md | testplan |
| 8 | 发布 / changelog / release | subskills/release.md | release |
| 9 | 运维手册 / oncall | subskills/runbook.md | runbook |
| 10 | 复盘 / postmortem / RCA | subskills/postmortem.md | postmortem |
| 11 | ADR / 为什么选 X | subskills/adr.md | decision |
| 12 | delta / change / 增量 | subskills/change.md | change |
What ships with it
12 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.
- subskills/adr.md 1.6 KB
- subskills/change.md 2.9 KB
- subskills/design.md 2.0 KB
- subskills/impl.md 5.6 KB
- subskills/plan.md 1.1 KB
- subskills/postmortem.md 2.1 KB
- subskills/prd.md 3.3 KB
- subskills/quick.md 721 B
- subskills/release.md 1.2 KB
- subskills/research.md 1.7 KB
- subskills/runbook.md 1.1 KB
- subskills/testplan.md 1.0 KB
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.
- 5d ago First seen · 144 lines · 92 tokens per session scan A 3c28e42a65d8
spec-manager is a skill published in the GitHub repository loki-ai-ch/spec-manager (5 stars, last pushed 1mo ago), licensed MIT. It adds 92 tokens to every session and 3,377 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
diagnose
Turn a bug symptom into a fix with a regression test that locks it down. If the user opens vague ("there's a bug", "/diagnose"), interview them one question at a time until you have enough to attempt a reproduction; if you still cannot reproduce, say so explicitly. Then drive the work through five phases — build a…
tdd
Drives feature work and bug fixes through a tight failing-test-first loop. Trigger when implementing, fixing, or refactoring behavior in a codebase that already has tests. Skip for spikes, visual-only edits, throwaway scripts, and generated files.
verify
Verify that a spec.md's Acceptance criteria are observably covered by green tests before opening a PR. Reads the spec, finds the tests that cover each AC, runs the suite to confirm everything green, and reports gaps (uncovered AC, skipped tests, missing edge-case coverage). When gaps exist, automatically iterates with…
spec-approach
Add a refined ## Approach section to an existing feature spec.md before TDD executes it. Reads the spec, explores the codebase to ground the approach in real modules and patterns, proposes 2–3 alternatives when they exist, validates with the user, then appends ## Approach to the same spec file. Trigger after…
feature
Kick off the full feature workflow — interview-to-spec, optional spec-approach, tdd, verify, then review — when starting a new feature from scratch. Thin orchestrator: confirms intent in one sentence, then hands off to the first stage and lets each downstream skill activate on its own. Trigger when the user says "new…
interview-to-spec
Conduct a focused interview to draft a spec.md for an upcoming task (the input step of SDD — spec-driven development). Walks through goal, behaviors, acceptance criteria, edge cases, and out-of-scope one branch at a time, then writes the spec to disk. If during the interview the scope turns out to be multiple…