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 Nitmi/baud-cli --skill baudgit clone --depth 1 https://github.com/Nitmi/baud-cliWrote 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/nitmi/baud-cli/baud)<a href="https://agentmods.dev/skills/nitmi/baud-cli/baud"><img src="https://agentmods.dev/badge/skills/nitmi/baud-cli/baud/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/nitmi/baud-cli/baud"><img src="https://agentmods.dev/badge/skills/nitmi/baud-cli/baud.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 | $0.00082 | $0.01026 |
| Opus 5.5 | $0.00033 | $0.00410 |
| Sonnet 5.5 | $0.00016 | $0.00205 |
| Haiku 4.5 | $0.00008 | $0.00103 |
Grade A, and why
baud 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 1mo 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 — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Baud
Use baud as the serial execution layer. Preserve evidence, minimize transmission, and gate physical actions behind verified device state.
Prepare
- Run
baud --versionbefore opening a port. - If the command is unavailable, report the missing prerequisite. When installation is authorized, use
uv tool install baud-cli; useuv run --directory <baud-cli-repo> baudonly when the source checkout is known. Do not silently replace it with an ad hocpyserialscript. - Read applicable
AGENTS.md, firmware documentation, and existing serial workflows to learn the baud rate, line ending, safe query commands, reset behavior, and physical risks. - Resolve the port from evidence. Do not assume a previous COM or tty name still identifies the same adapter.
Choose The Least Invasive Command
Follow this escalation order:
- Run
baud list --jsonto enumerate ports and USB metadata. - Run
baud monitor --port <port> --duration <seconds> --jsonto observe boot output without transmitting. - Run
baud probe --port <port> --commands <safe queries> --endings crlf lf --jsononly after confirming the probe commands are harmless for the device. - Run
baud sendfor one bounded query with explicit expectations. - Run
baud run <workflow.yaml> --jsonfor multi-step configuration, observation, or hardware action.
Use --json for a bounded command whose final result drives the next decision. Use --jsonl when event ordering or streamed evidence matters. Keep automatic log artifacts enabled for real hardware work.
Apply Safety Gates
- Start with read-only observation and status queries.
- Treat DTR and RTS as physical control signals that may reset or reconfigure a device. Keep both false unless device documentation requires otherwise.
- Confirm both communication directions before sending a state-changing command. A boot banner proves only device-to-host traffic.
- Read configuration back from the device; do not treat ACK, echoed input, or a successful write as proof that hardware accepted the state.
- Mark motion, heating, injection, erase, flash-save, power switching, and similar actions as
dangerous: truein workflows. Require successful earlier verification steps withrequires. - Begin physical testing with one cycle, the lowest safe setpoint, unloaded mechanics, or another documented low-risk condition.
- Stop on a busy, disappearing, or unidentified port. Do not kill an unknown process or continue on a different port without evidence.
What ships with it
3 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.
- 1mo ago First seen · 71 lines · 82 tokens per session scan A dab6a7b478ea
baud is a skill published in the GitHub repository Nitmi/baud-cli (5 stars, last pushed 1mo ago), licensed MIT. It adds 82 tokens to every session and 1,026 once invoked, about $0.0003 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 2026-08-31.
Other skills, from other repositories
serial-debug
CLI-first serial port debugging and macro automation with serial-mcp-server for Codex and Claude Code. Use when working with UART or USB-serial devices, listing serial ports, probing STM32/Arduino/ESP32 boards, writing or reading serial data, running JSON serial macros, controlling RTS/DTR, or using serial-mcp-server…
llama-cpp
Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
volt-ota
Produce a complete OTA update system design — partition layout, update flow, rollback conditions, validation checks, fleet management approach, failure modes and recovery. Use when asked about "OTA updates", "firmware updates over the air", "how do I update devices in the field", "OTA strategy", or "remote firmware…
volt-firmware
Produce a complete firmware architecture spec for a described device — layer diagram, module responsibilities, HAL interface definitions, key state machines, RTOS decision. Use when asked to "design firmware architecture", "plan embedded firmware", "architect an IoT device", "how should I structure this firmware", or…
volt-driver
Build a device driver or protocol handler — I2C sensors, BLE services, MQTT clients, SPI peripherals with interrupt-driven I/O and clean HAL abstraction. Use when asked to "write a driver", "I2C device", "BLE service", "MQTT client", or "sensor integration".
volt-power
Power management audit — analyze sleep modes, wake sources, power state machines, radio duty cycles, and battery life estimates. Use when asked to "audit power usage", "optimize battery life", "review power management", "why is my battery draining", "power budget analysis", or "sleep mode review".