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/bbl21/cst-runtime-cli/cst-runtime-optimizationnpx skills add bbl21/cst-runtime-cli --skill cst-runtime-optimizationgit clone --depth 1 https://github.com/bbl21/cst-runtime-cliWhat 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 | $0.00076 | $0.04877 |
| Opus 5 | $0.00038 | $0.02439 |
| Sonnet 5 | $0.00015 | $0.00975 |
| Haiku 4.5 | $0.00008 | $0.00488 |
Grade A, and why
cst-runtime-optimization 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.
How it starts
The opening of the file, as written. The whole thing — 386 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CST Runtime 优化 Skill
优化流程速查
Phase 0: 准备
task 确认(工程/目标/参数/约束) → inspect-project(含category)
[agent: 选几何参数 → 用户确认 → 翻译为objective spec]
Phase 1: 探针(参数≥4强制)
run-probe-phase --objective <spec>
[agent: 看top_params/edge_hit → 选参数/调范围/定算法]
Phase 2: 优化循环
loop:
run-optimization-step --objective <spec>
[agent: objective达标? → break / ask_study_failed? → 换优化器]
Phase 3: 收尾
generate-report → 灵敏度验证 → 经验→docs/
[agent: 收敛质量判断 + 结论沉淀]
关键决策点：
| 阶段 | 节点 | agent 行动 |
|---|---|---|
| 0 | inspect-project | 筛选 category=geometry 的参数作为候选 |
| 0 | task确认 | 向用户确认5项：工程类型、objective函数、目标阈值、参数列表、几何约束 |
| 1 | run-probe-phase 返回 | main_effects_normalized > 0.1 保留，edge_hit=true 扩范围 |
| 1 | 算法选择 | 参数少+线性=TPE，多+交互复杂=CMA-ES，默认 suggested_algorithm |
| 2 | 每轮返回 | objective_value vs 目标阈值 → 达标break |
| 2 | ask_study_failed | 未达标且 study 认为收敛 → 换优化器(最多一次) |
| 3 | 收尾 | 灵敏度验证 + 经验入库 |
定位
本 Skill 专注 CST 参数优化闭环，依赖 cst-runtime-cli 提供底层基础设施。
- 不携带
scripts/或cst_runtime/源码；所有 CLI 调用走 base skill。 - 负责定义优化迭代流程、早停判断、参数策略、数据导出和报告生成。
- 所有生产任务使用标准
tasks/task_xxx_slug/runs/run_xxx/{projects,exports,logs,stages,analysis}结构。
依赖声明
本 Skill 不实现 CST 操作，以下工具全部由 cst-runtime-cli 提供。所有 CLI 调用通过工作区的 -m cst_runtime 入口执行。
| 职责 | CLI 工具 |
|---|---|
| run 创建 | prepare-run、get-run-context |
| 审计 | record-stage、update-status |
| 进程/session | cst-session-open、cst-session-close、cst-session-quit |
| 参数 | list-parameters、change-parameter |
| 仿真 | start-simulation-async、wait-simulation |
| 结果导出 | export-run-results（统一导出 S11+2D+远场） |
| 结果显示 | generate-report（生成综合报告） |
| 探针阶段 | run-probe-phase（一键探针：设计→仿真→分析→注入 study） |
| 优化迭代 | run-optimization-step（一步迭代：ask→改参→仿真→tell，agent 判断早停） |
| Objective 函数 | 内置在管道中 — s11_min_db, s11_at_freq, gain_max, bandwidth, expression |
Objective 函数系统
优化目标不再硬编码为 S11 min dB。通过在管道工具中传入 --objective <spec> 指定目标函数：
| 类型 | 格式 | 示例场景 |
|---|---|---|
| S11 全局最小值 | {"type": "s11_min_db"} |
默认，最小化反射 |
| 指定频率 S11 | {"type": "s11_at_freq", "freq": 2.4} |
2.4GHz 阻抗匹配 |
| 最大增益 | {"type": "gain_max", "port": 1} |
优化天线增益 |
| 带宽 | {"type": "bandwidth", "below_db": -10} |
S11<-10dB 带宽最大化 |
| 自定义表达式 | {"type": "expression", "expr": "min(s11_db)"} |
灵活组合 |
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 First seen · 386 lines · 76 tokens per session scan A d8f2660b8072
cst-runtime-optimization is a skill published in the GitHub repository bbl21/cst-runtime-cli (25 stars, last pushed 2mo ago), licensed MIT. It adds 76 tokens to every session and 4,877 once invoked, about $0.0004 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
musicbox
Use when the user wants to play/pause/skip music, control volume, seek, search songs/playlists, query NetEase Music data, or operate NetEase MusicBox. Drives MusicBox through the musicbox CLI + daemon; never simulates terminal keypresses to the curses TUI.
codexspec:review-code
将所选变更作为严格缺陷门禁进行审查，或使用 --audit 审计路径.
codexspec:distill
Distill reusable, cross-feature knowledge from an interaction into the project profile.
codexspec:implement-tasks
执行实现任务，支持条件 TDD 工作流（代码使用 TDD，文档/配置直接实现）.
codexspec:constitution
通过交互式或提供的原则输入创建或更新项目宪法，确保所有依赖模板保持同步。.
codexspec:pr
基于 git diff 和可选的 spec.md 集成生成结构化的 Pull Request (GitHub) 或 Merge Request (GitLab) 描述.