メニュー
Run the protocol fuzzer (`ssgg fuzz`) against a connected keyboard with safety guardrails, capture responses, and fold results back into docs/development/ notes.
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
Run steelseriesgg-rs's benchmark binaries before and after a change and report the timing delta.
Run the full local CI matrix for steelseriesgg-rs, matching .github/workflows/ci.yml exactly. Use this whenever the user asks to run CI, check all feature variants, run the full test suite, verify before pushing, or run clippy/fmt/tests across features. Invoke proactively before any push or PR review.
Rust conventions for steelseriesgg-rs — error handling, unsafe, feature-gated commands. Use when writing or refactoring any .rs file or Cargo.toml.
Use when user wants to Find and remove genuine dead code using deadcode + vulture. Filters known false positives (dataclass fields, protocol methods, dunder hooks). Run before a cleanup commit.
HID protocol constraints for steelseriesgg-rs device code. Automatically load when editing any file under src/devices/. Contains hard rules about HidReportBuilder usage, unsafe block documentation, and experimental command code labeling.
| name | fuzz-runner |
| description | Run the protocol fuzzer (`ssgg fuzz`) against a connected keyboard with safety guardrails, capture responses, and fold results back into docs/development/ notes. |
| user-invocable | true |
| disable-model-invocation | true |
Run src/devices/fuzz.rs (Commands::Fuzz in src/main.rs, hidden from --help) against real hardware. This sends arbitrary command bytes to a physical device — it can brick it or cause unexpected behavior. Only run this when the user explicitly asks to fuzz a device.
ssgg list (or equivalent device-discovery command) and confirm the user is aware this targets that specific device.0x00..0xFF sweep unless they ask for it. Narrower ranges (e.g. probing around a known command like 0x2D ActuationControl) are safer and faster to interpret.cargo run --bin ssgg -- fuzz --start <hex> --end <hex> --delay <ms>
--delay defaults to 100ms between commands; don't lower it without reason — faster sweeps give firmware less time to recover between probes.Sent/Failed per command and Received Response when the device replies. Capture full stdout — the response pattern (which commands ACK vs. time out) is the actual research signal.0x23, 0x40, 0x2D — see .claude/rules/experimental-protocol.md), note that as a hypothesis, not a confirmation — promoting a code from experimental to confirmed requires the process in that rule file (device model + date in the commit).docs/development/ (check existing files there for the format other reverse-engineering notes use) rather than creating a new top-level doc.