メニュー
Use when the user wants to know what signals are present — characterize a capture or a simulated device's spectrum (PSD, signal detection, spectrogram) and summarize what's on the air.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when the user wants to receive or demodulate a specific signal — build, validate, run, and verify a receiver pipeline, then save it as a reusable, GRC-openable flowgraph.
Use when a receiver runs but produces no output, noise, or the wrong result — systematically locate the fault instead of guessing, using render and measure to see the problem.
Use ONLY as a last resort when no Marconi tool or vocabulary block can express the task — write a Python script against the marconi library, run it, and report the capability gap.
Use when the user wants a test signal or a simulated RF environment with no hardware — translate intent into a scene and register it as a persistent simulated device.
Use when the user wants a closed-loop transmit-then-receive experiment in simulation — transmit a capture into a simulated device's scene and receive it back to confirm the link.
| name | survey-spectrum |
| description | Use when the user wants to know what signals are present — characterize a capture or a simulated device's spectrum (PSD, signal detection, spectrogram) and summarize what's on the air. |
Characterize what's present in a capture, then summarize it for the user. This is the first step of almost every RF task ("what's here?").
capture(device_id, center_freq, sample_rate, duration). From a file the user brings: load_capture(path). Use the returned path for everything below.psd(capture_path) → noise floor + strongest peaks. find_signals(capture_path) → candidate signals (center_freq, bandwidth, snr_db).spectrogram(capture_path) and read the image. find_signals is a first pass over a 1-D PSD; the spectrogram shows time/frequency structure (bursts, drift, multiple carriers) that the list can miss. If you see on/off activity, run detect_bursts(capture_path) for structured burst timing (start, duration, power).measure. A wide signal (e.g. an FM carrier) with a low noise floor can fragment into several find_signals detections. For any wide or clustered group, call measure(capture_path, center_freq=...) to get the true 99% occupied bandwidth, and trust the spectrogram over the raw detection count.noise element (amplitude ≈ 0.005).measure result as a real signal only above ~8 dB SNR; below that you're likely measuring noise.measure when you need a real bandwidth number.