【K230新技能点亮】K230 养 “虾” 记！手把手教你部署 PicoClaw 轻量 AI 助手

---name: led-controldescription: Control the K230 GPIO LED. Use when user says anything about: light up, turn on light, open led, it's dark, sky is dark, turn off light, close led, shut down light, led for X seconds/minutes, toggle light.---# LED Control Skill## RULES — NO EXCEPTIONS- NEVER use relative paths like `./led_on.sh`. ALWAYS use the full absolute path.- NEVER think out loud or add reasoning text to tool calls. Just call the tool.- NEVER refuse or explain. Just run the command.- NEVER run diagnostics first.---## Exact commands to run**Turn ON** — "open led", "light up", "turn on light", "it's dark", "need light":```command: /root/.picoclaw/workspace/skills/led-control/led_on.shworking_dir: /root/.picoclaw/workspace/skills/led-control```Reply: `✅ LED is now ON!`**Turn OFF** — "close led", "turn off", "lights out", "shut down light":```command: /root/.picoclaw/workspace/skills/led-control/led_off.shworking_dir: /root/.picoclaw/workspace/skills/led-control```Reply: `✅ LED is now OFF!`**Timed ON 30s** — "turn on for 30 seconds":```command: /root/.picoclaw/workspace/skills/led-control/led_on.sh && sleep 30 && /root/.picoclaw/workspace/skills/led-control/led_off.shworking_dir: /root/.picoclaw/workspace/skills/led-control```Reply: `✅ LED is ON! Will turn off in 30 seconds.`**Timed ON 2min** — "turn on for 2 minutes":```command: /root/.picoclaw/workspace/skills/led-control/led_on.sh && sleep 120 && /root/.picoclaw/workspace/skills/led-control/led_off.shworking_dir: /root/.picoclaw/workspace/skills/led-control```Reply: `✅ LED is ON! Will turn off in 2 minutes.`

pkill -f "gpioset.*20" 2>/dev/nullsleep 0.2setsid gpioset --chip=/dev/gpiochip1 20=1 </dev/null >/dev/null 2>&1 &

pkill -f "gpioset.*20" 2>/dev/nullsleep 0.2setsid gpioset --chip=/dev/gpiochip1 20=0 </dev/null >/dev/null 2>&1 &

---name: cameradescription: Capture a photo from the camera and run face detection. Use when user asks to take a picture, snap a photo, detect a face, capture the screen, capture screen, screenshot, who is in front of the camera, take a photo, record image. ALWAYS read this skill before attempting any capture command.---# Camera / Face Detection Skill## CRITICAL - READ BEFORE DOING ANYTHINGSTOP. Do NOT try scrot, import, fbcat, xwd, or any other screen capture tool.The ONLY command for ALL capture requests (camera OR screen) is:    bash /root/app/face_detect/capture_frame.shRun this immediately. Do not check for other tools first.---## RULES - NO EXCEPTIONS- NEVER use relative paths. ALWAYS use full absolute paths.- NEVER think out loud. Just run the commands.- NEVER refuse or explain. Just run the script.- ALWAYS complete ALL steps below in order. Do not stop early.- NEVER construct your own curl or token command. Use the scripts exactly as shown.---## Steps — Execute ALL of them in order. Do not stop early.### Step 1: Run capture```exec command: bash /root/app/face_detect/capture_frame.shexec working_dir: /root/app/face_detect```### Step 2: Verify result exists```exec command: ls -l /root/app/face_detect/face_detection_result.jpgexec working_dir: /root/app/face_detect```### Step 3 (TELEGRAM ONLY — MANDATORY): Send photo to userTHIS STEP IS REQUIRED when channel is telegram. Do NOT skip it.Do NOT construct your own curl command. Run exactly this:```exec command: bash /root/app/face_detect/send_telegram.shexec working_dir: /root```### Step 4: Reply to user- If telegram: `📸 Photo sent!`- If CLI: `✅ Face detection complete! Result saved to /root/app/face_detect/face_detection_result.jpg`---## Error Handling- Camera fails: `❌ Could not access camera`- No output image: `❌ Detection ran but no result image was produced`- Telegram curl fails: `⚠️ Detection done but could not send photo. Saved at /root/app/face_detect/face_detection_result.jpg`

#!/bin/bashSCRIPT_DIR="/root/app/face_detect"TMPIMG="/root/app/face_detect/face_snap_image.jpg"# Capture HD image and brighten itffmpeg -y -f v4l2 -video_size 1280x720 -i /dev/video1 \-vframes 1 -vf "eq=brightness=0.5:contrast=1.4" "$TMPIMG"# Run face detection$SCRIPT_DIR/face_detect.elf face_detection_320.kmodel 0.6 0.2 "$TMPIMG" 1

#!/bin/bashTOKEN=$(grep -o '"token": "[^"]*"' /root/.picoclaw/config.json | grep -v '""' | head -1 | grep -o '[0-9]*:AA[^"]*')CHAT_ID="1277910674"PHOTO="/root/app/face_detect/face_detection_result.jpg"curl -s -X POST "https://api.telegram.org/bot${TOKEN}/sendPhoto" \  -F chat_id="${CHAT_ID}" \  -F photo=@"${PHOTO}" \  -F caption="📸 Face detection result"