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.
git clone --depth 1 https://github.com/PsChina/deepseek-as-subagentWrote 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/commands/pschina/deepseek-as-subagent/ds)<a href="https://agentmods.dev/commands/pschina/deepseek-as-subagent/ds"><img src="https://agentmods.dev/badge/commands/pschina/deepseek-as-subagent/ds/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/commands/pschina/deepseek-as-subagent/ds"><img src="https://agentmods.dev/badge/commands/pschina/deepseek-as-subagent/ds.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 | $0.00026 | $0.00561 |
| Opus 5.5 | $0.00010 | $0.00224 |
| Sonnet 5.5 | $0.00005 | $0.00112 |
| Haiku 4.5 | $0.00003 | $0.00056 |
Grade A, and why
ds 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 2d 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.
What it actually says
/ds — Delegate to DeepSeek
把后面跟的任务强制派给 DeepSeek 处理，绕过 Claude 的自动决策（"该派 / 不该派"）。
/ds 固定调用完整 coding API；明确只做静态文件分析的任务应由主 Agent 选择
delegate_to_deepseek_readonly。
模型选择遵循 delegate-to-deepseek skill：Flash 是强前沿 coding-agent 档，粗略处于高端 Sonnet 到 Opus 竞争区间，默认用于大多数 coding 任务；Pro 目标约为 Opus / Sol 级，用于复杂 debugging、架构推理、困难跨文件任务或 Flash 已失败后的升级。
注意：截至 2026-10-22，DeepSeek 官方说明 deepseek-v4-pro 在 V4.1-Pro 上线前会路由到 V4.1-Flash。如果配置的 Pro 槽位仍使用该别名，就不会产生实际能力升级。具体路由和任务边界见 skill 第 3 节。
你要做的
-
按
delegate-to-deepseekskill 的准则准备task、context和必要时的model：- 用 Glob / LS 收集涉及的文件路径
- 摘要项目约定（命名规则、输出 schema、边界）
- 写明成功标准
- 普通任务保持默认 Flash；困难任务才显式选择 Pro
-
调用
mcp__deepseek__delegate_to_deepseek工具，把用户的请求当作 task 传入：
用户输入: $ARGUMENTS
- 工具返回后必须验证：
- Read 抽样产物文件
- 检查数量 / schema sanity
- 失败按 skill 的 fallback 策略处理
不要做的
- ❌ 不要在调用前问用户"你确定要派吗" —— 用户敲
/ds已经是明确指令 - ❌ 不要因为 Pro 可用就默认选择 Pro
- ❌ 不要把真实 provider 模型名传给
model参数 - ❌ 不要在工具返回 ERROR 时直接放弃 —— 按 skill 的 fallback 策略重试或接管
- ❌ 不要把 API key / 凭证塞进 task / context
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.
- 2d ago Changed 4804583d50cc
- 1mo ago First seen · 41 lines · 26 tokens per session scan A 9635bee0de64
ds is a command published in the GitHub repository PsChina/deepseek-as-subagent (52 stars, last pushed 3d ago), licensed MIT. It adds 26 tokens to every session and 561 once invoked, about $0.0001 per session on Opus 5.5. 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 commands, from other repositories
kmp-setup-agents
Initialize .agents/ in an existing KMP project so the team gets agent-driven workflows without running the full scaffold.
kmp-audit-screenshots
Analyze Roborazzi golden screenshots for design consistency.
kmp-migrate-to-shadcn
Migrates an existing project from the owned kmp-compose-design-system (App components) to the published shadcn-compose library (Shadcn components).
kmp-audit-design-visual
Runs a visual-only audit of your app's Roborazzi screenshots using Claude vision.
kmp-modify-skill
Modify an existing skill safely: add or update a section, fix a pattern, bump library versions, or resolve a flagged quality gap.
kmp-new-skill
Scaffold a new SKILL.md from scratch, following every structural rule enforced by auditskillsrepo.py and scanskillissues.py.