Borrowing it
Nothing to install: this file belongs to SensorsIot/Embedded-AI-Harness. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/SensorsIot/Embedded-AI-Harness/main/.claude/skills/testbench-install/SKILL.mdgit clone --depth 1 https://github.com/SensorsIot/Embedded-AI-HarnessWrote 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/sensorsiot/embedded-ai-harness/testbench-install)<a href="https://agentmods.dev/skills/sensorsiot/embedded-ai-harness/testbench-install"><img src="https://agentmods.dev/badge/skills/sensorsiot/embedded-ai-harness/testbench-install/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/sensorsiot/embedded-ai-harness/testbench-install"><img src="https://agentmods.dev/badge/skills/sensorsiot/embedded-ai-harness/testbench-install.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 6 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 107 Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.Fix: Limit tool chaining depth and validate the output of each tool before passing it to the next. Require explicit user approval for multi-step chains.
- medium Privilege Escalation · line 23 Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.Fix: Avoid sudo/root unless strictly required. Prefer least-privilege patterns. If elevation is needed, document the justification and scope.
- medium Privilege Escalation · line 32 Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.Fix: Avoid sudo/root unless strictly required. Prefer least-privilege patterns. If elevation is needed, document the justification and scope.
- medium Privilege Escalation · line 71 Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.Fix: Avoid sudo/root unless strictly required. Prefer least-privilege patterns. If elevation is needed, document the justification and scope.
- medium Privilege Escalation · line 107 Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.Fix: Avoid sudo/root unless strictly required. Prefer least-privilege patterns. If elevation is needed, document the justification and scope.
- medium Rogue Agent · line 39 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
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 | $0.00143 | $0.02052 |
| Opus 5 | $0.00072 | $0.01026 |
| Sonnet 5 | $0.00029 | $0.00410 |
| Haiku 4.5 | $0.00014 | $0.00205 |
Grade B, and why
testbench-install scanned grade B with 2 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
| Bench works, want the latest code + no system changes | `sudo bash install.sh --update` | Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
finished; `curl /api/info` is the only proof. How it starts
The opening of the file, as written. The whole thing — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Setting up and deploying the testbench
The one fact that explains most confusion: the service runs from
/usr/local/bin/, not from the git checkout. Editing files in the repo on the
Pi — or on your laptop — changes nothing until they are copied across and the
service is restarted. Every "my fix didn't work" report traces back to this.
Full operator procedure lives in
docs/Harness-User-Manual.md
§2. This skill is the working procedure plus the things that only bite when you
actually do it.
Pick the operation
| Situation | Do this |
|---|---|
| New Pi, blank SD card | Fresh install below |
| Bench works, want the latest code + no system changes | sudo bash install.sh --update |
| Changed one module, want it live now | Single-file deploy below |
Changed a config default in pi/config/ |
Fresh-install path won't overwrite it — see Config files are never overwritten |
Fresh install
git clone https://github.com/SensorsIot/Embedded-AI-Harness.git
cd Embedded-AI-Harness/pi
sudo bash install.sh
install.sh is idempotent and does eight things: apt packages, standing down
the services it manages dynamically (hostapd is masked; dnsmasq and
mosquitto are disabled — the portal starts them itself, so leaving them
enabled fights it), directories, the Python modules into /usr/local/bin/,
helper scripts, config defaults, systemd + udev rules, then enable and start.
It fetches openocd-esp32 from GitHub releases and installs it as
/usr/local/bin/openocd-esp32, alongside — not replacing — Debian's openocd.
It runs under set -e, and the script copy comes before systemd and udev. So
a single missing file aborts the install after the packages are in but before the
service exists, and the output ends on a bare cp: cannot stat with no summary.
Read the last line rather than assuming a long successful-looking log means it
finished; curl /api/info is the only proof.
The API contract is FSD Appendix D. Every endpoint the bench serves is listed there with its request and response shape. Read it rather than the portal source: the source has carried a dead duplicate handler whose documented form the live one rejects, so guessing from code has already cost a debugging cycle.
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 · 178 lines · 0 tokens per session scan B 6c9bcf6a4bfa
testbench-install is a skill published in the GitHub repository SensorsIot/Embedded-AI-Harness (173 stars, last pushed 29d ago), licensed MIT. It adds 143 tokens to every session and 2,052 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it B with 2 findings (asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
m5stack-cap-lora-1262
Hardware reference and firmware helper for the M5Stack Cap LoRa-1262 (SKU U214) — a snap-on cap for the Cardputer Adv (K132-Adv) and CardputerZero that carries a Semtech SX1262 sub-GHz LoRa radio (868–923 MHz, +22 dBm TX, external RP-SMA antenna) and an ATGM336H-6N GNSS receiver (GPS/QZSS/BeiDou/Galileo/GLONASS, UART…
new-device-skill
Build or update an M5Stack hardware skill in this marketplace — a Controller (a board that runs firmware, like Core2 or AtomS3), a Unit (a peripheral you drive from a Controller, like a ToF sensor or a relay), or a Chip (an Espressif SoC capability layer, like esp32-c6). Use whenever adding a new board/unit/chip to…
m5stack-tab5
Hardware reference and development helper for the M5Stack Tab5 (product code C145) — an ESP32-P4-based 5" touchscreen IoT/industrial terminal with an ESP32-C6 wireless co-processor, MIPI-DSI display, MIPI-CSI camera, ES8388 audio codec, BMI270 IMU, RX8130CE RTC, RS485, and a removable NP-F550 battery. Use this skill…
m5stack-core2
Hardware reference and development helper for the M5Stack Core2 family — a 2.0" touchscreen ESP32 (classic, Xtensa LX6) Controller built around an AXP192 power management IC, ILI9342C display, FT6336U capacitive touch, BM8563 RTC, and (on original/1.1/1.3 revisions) an MPU6886 or BMI270 IMU. Covers the plain Core2…
esp32-c6
Chip-level ESP-IDF capability reference for the ESP32-C6 SoC (single-core RISC-V HP core + RISC-V LP core, WiFi 6 + BLE 5.3 + Thread/Zigbee) — what the chip can do, distinct from any board's wiring. Use when a user wants to exploit ESP32-C6 hardware — WiFi 6/BLE/802.15.4 radio coexistence (Thread Border Router, Zigbee…
esp32
Chip-level ESP-IDF capability reference for the classic ESP32 SoC (dual-core Xtensa LX6) — what the chip can do, distinct from board wiring. Covers the ESP32-D0WDQ6-V3 / ESP32-D0WD-V3 / ESP32-D0WDR2-V3 die family used in ESP32-WROOM-32 and ESP32-WROVER modules. Use for RMT (IR/WS2812), LEDC (PWM), I2S (audio/parallel…