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 hxt9805/cadence --skill project-discussgit clone --depth 1 https://github.com/hxt9805/cadenceWrote 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/hxt9805/cadence/project-discuss)<a href="https://agentmods.dev/skills/hxt9805/cadence/project-discuss"><img src="https://agentmods.dev/badge/skills/hxt9805/cadence/project-discuss/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/hxt9805/cadence/project-discuss"><img src="https://agentmods.dev/badge/skills/hxt9805/cadence/project-discuss.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00216 | $0.03801 |
| Opus 5 | $0.00108 | $0.01901 |
| Sonnet 5 | $0.00043 | $0.00760 |
| Haiku 4.5 | $0.00022 | $0.00380 |
Grade A, and why
project-discuss 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 9d 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 — 248 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Project Discuss
L1 协议入口：管理项目讨论、记录决策、处理查询、记录 incidents。
本 SKILL.md 是动作骨架 + Phase 化导航。L0 借口反驳 4 项（#1/#2/#4/#5）+ 三段式判定 happy path 已在 cadence-bootstrap/SKILL.md inline；细则单一权威源在 references/，按需读取：
references/recording-protocol.md— 记 阶段细则 + 信息密度正反例 + entry schema + incidents 附录（v0.5 合并 incident-handling.md）references/recording-fidelity.md— 通用语义保真：三段式判定、Light/Standard/High、承接短句回溯、冷启动检查references/query-behavior.md— 查询行为 + 4 trigger 主动重读 + 文档可信度 L1-L4（v0.5 合并 doc-reliability-protocol.md）references/harness-adapters.md— Codexspawn_agent+ OpenCode 工具映射（v0.5 合并 codex-tools.md + opencode-tools.md）
§ 1. 讨论开始前（必做）
第 1 步：检查 cadence 骨架
读 cadence/_INDEX.md：
- 不存在 → 告知用户："看起来这是新项目，还没初始化 cadence。我可以先跑
/cadence-init建好骨架，然后继续讨论。要开始吗？" - 存在 → 继续第 2 步
第 2 步：建立全局上下文
- 读
cadence/_INDEX.md（纯索引，< 800 tokens）—— 项目简述 / 话题词典 / 快速导航 - Session 内第一次需要活跃状态时，读
cadence/_ACTIVE.md（~1800 tokens）—— 活跃决策 / 待决 / TODO / 最近讨论 - 用户提到的话题在词典里 → 读指向的文档；用户问历史话题（> 14 天前）→ 读
_INDEX-HISTORY.md - 不全量加载：默认不重读，4 trigger 时重读（详
references/query-behavior.md）
§ 2. Phase 化协议骨架
完整细则见 references/recording-protocol.md。
Phase 1 记 — 三段式判定 / 落点 / 告知 / 铁律 / 质量自检
- 承接：用户明确或隐含确认了候选方案
- 增量：该方案改变了可持续影响后续工作的目标、范围、规则、顺序、边界、状态、风险、依赖、否决项或下一步
- 深度：按通用影响维度选择 Light / Standard / High；不是所有 entry 都套同一个 minimum
- 落点：
cadence/streaming/<YYYY-MM-DD>-<topic-slug>.md - 铁律：append-only（不修改已有 entry；撤回 = append tombstone）
- 告知：一行
📝 已记：<摘要> → <path>
用户只说“可以”“认可”“方案二”或“按你推荐的来”时，先回溯它明确指向的最近完整方案。记录对象是被承接方案的持久语义，不是承接短句本身；范围明确时自动提取并记录，只有指向多个候选方案而无法可靠判断时才询问用户。
★ profile-aware 质量自检 checklist（扩展 L0 §5c 简版）：
所有 profile 都先检查：
☐ chosen — 选了什么，适用边界是什么？
☐ context — 为什么此时需要决定？
☐ status — accepted / pending / superseded？
☐ provenance — explicit / synthesized / inferred 是否与内容一致？
- Light：
context + chosen - Standard：Light +
rationale；真实讨论过时再写options/rejected/dependencies/open_questions - High：
context + chosen + rationale+ 适用的通用语义槽；不适用但容易被误解为遗漏的槽写入not_applicable
What ships with it
7 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.
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.
- 9d ago First seen · 248 lines · 216 tokens per session scan A 79dd8c865d28
project-discuss is a skill published in the GitHub repository hxt9805/cadence (5 stars, last pushed 1mo ago), licensed MIT. It adds 216 tokens to every session and 3,801 once invoked, about $0.0011 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
flow-next-plan
Plan a feature into a flow-next spec with tasks in .flow/. Use when asked to plan, spec out, or break down work (fn-N ids).
flow-next-tracker-sync
Project a flow-next spec to a tracker issue (Linear, GitHub, GitLab, Jira) and reconcile two-way. Use when asked to sync to a tracker. NOT plan-sync.
flow-next
Manage .flow/ tasks and specs. Use for show or list tasks, task status, what is ready, show fn-N. NOT for planning or executing (use the plan and work skills).
kanban
Use when a prompt contains a new bug/feature/change not yet tracked, or when an item's status changes — capture and maintain it in the project's beads tracker (Inbox→Backlog with Definition of Ready, priorities, the pipe-separated note format).
subtask-update
Update subtask with fixes or new steps. Two trigger modes: (A) Auto-called by autoworker:gate-check on FAIL — reads FAIL info, adds supplementary steps. (B) User calls with a finding/bug during testing — diagnoses, fixes, updates subtask. Both modes end by invoking autoworker:dispatch to re-enter the loop.
checkpoint
Record progress to subtask: Phase completion (from autoworker:code) or test results (from autoworker:test). Auto-detects upstream type from conversation context. Ends by calling autoworker:dispatch.