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/superjj007/csswitch/csswitch-external-skill-toolsnpx skills add SuperJJ007/CSSwitch --skill csswitch-external-skill-toolsgit clone --depth 1 https://github.com/SuperJJ007/CSSwitchWhat 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 | $0.00054 | $0.02282 |
| Opus 5 | $0.00027 | $0.01141 |
| Sonnet 5 | $0.00011 | $0.00456 |
| Haiku 4.5 | $0.00005 | $0.00228 |
Grade A, and why
csswitch-external-skill-tools 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 yesterday.
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 — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CSSwitch external Skill tools
This Skill only routes external Skill operations. It is not an authoring or
publishing workflow. For every request in scope, use the CSSwitch external Skill
connector and its matching tool. Do not load customize.
Install or import
- Accept a public GitHub repository root, ref root, Plugin/Skill collection, or exact Skill directory URL. Preserve it exactly. If the user supplied only a name, ask for the URL; do not search for or guess a source.
- Load the generated connector Skill
mcp-csswitch-skill-installer. - Follow that connector's API and call
install_external_skillwith the exactsource_url. The equivalent REPL call is:
host.mcp(
"csswitch-skill-installer",
"install_external_skill",
source_url=source_url,
)
Never call host.skills.edit or host.skills.publish as a fallback.
Never use Add Skill ZIP or marketplace.importSkills as a fallback.
Never download the Skill yourself, use shell/Python filesystem APIs, use a
GitHub credential, or pass staged files to CSSwitch. The CSSwitch host owns URL
resolution, one-archive download, validation, atomic commit, and OPERON
attachment. Repository and collection URLs may install a Nature-like bundle;
this installs only its Skill collection and support resources, not hooks, MCP,
agents, or a complete Claude Plugin runtime.
Exact-plan confirmation for one fixed-commit Skill
When the user asks to review and explicitly confirm a single-Skill install, and
the supplied source_url already contains a 40-hex GitHub commit SHA, first
request a plan without downloading or installing anything into Science:
planned = host.mcp(
"csswitch-skill-installer",
"install_external_skill",
source_url=source_url,
confirmation={"schema_version": 1, "action": "plan"},
)
If it returns CONFIRMATION_REQUIRED, show the user its exact source, complete
effects, plan_digest, expiry, and any degradation. Host access approval is
not the user's confirmation of those effects. Only after an explicit approval,
call once with the returned values unchanged:
What ships with it
4 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.
- yesterday First seen · 225 lines · 54 tokens per session scan A cebabd2b483f
csswitch-external-skill-tools is a skill published in the GitHub repository SuperJJ007/CSSwitch (452 stars, last pushed 9d ago), licensed MIT. It adds 54 tokens to every session and 2,282 once invoked, about $0.0003 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
alphafold2
Predict protein structure for monomers and multimers with AlphaFold2 via the ColabFold runner (Mirdita et al. 2022, github.com/sokrypton/ColabFold; AlphaFold2 Jumper et al. 2021). Reach for this skill to fold a sequence or complex with the AF2/AF2-Multimer evoformer, to validate designed sequences by self-consistency…
figure-library
Build, review, search, select, materialize, and safely publish immutable scientific-figure references from Local Published, bundled Community, FigureYa, and explicitly trusted personal Providers.
vision-multimodal
给 DeepSeek 这类纯文本模型补上「眼睛和耳朵」的完整视觉/听觉能力 —— 读图、OCR 取字、物体检测定位、视频理解、语音转写、多说话人分离、音乐分析、文档/PDF 可视化、截图直读、图像分割。双通道：默认走 MCP 工具（通义千问），也可用 vision.py 接任意 OpenAI 兼容供应商。.
motif-for-claude-science
Opens bounded sequence data in the connected Motif workbench or creates a self-contained Motif HTML artifact for visual review.
remote-compute
Use when the user asks to run, submit, monitor, or cancel a job on a remote machine over SSH — their own GPU/CPU server, a workstation, or a Slurm cluster ("the cluster", a login node, "my 3090 box", "the compute server"). Picks a saved machine, runs the work directly over SSH (or via Slurm when present), tracks it…
modal-run
Use when the user asks to run heavy or GPU work on Modal (the cloud compute platform) — writing a Modal function in the workspace, running it with the user's own modal CLI + token, and bringing results back. Data-to-compute for jobs too big for the laptop, without a Slurm cluster.