dd

dd — Stop Dictation

Self-contained fast path for /dictate stop — do NOT invoke the dictate skill (that would cost an extra round-trip).

Run ONE Bash call (stop plays the falling tone and waits for the transcript flush):

SET_COPILOT_DIR="$PWD/.set/copilot/${CLAUDE_CODE_SESSION_ID:-shared}" npx set-copilot stop --print

SET_COPILOT_DIR must match the one /ds used — that directory holds this session's PID file and transcript. Without it, stop would look in the global runtime dir and could kill another session's capture.

--print emits the transcript and archives it in one step, so it is handed over exactly once: a second /dd prints nothing instead of replaying the previous dictation as if it were freshly spoken. It also prints a transcript left behind by a capture that already self-stopped on its time limit.

The output is plain text: the dictated words, already reassembled into sentences. There is nothing to parse and nothing to join.

⚠ Lines prefixed [set-copilot] are the tooling talking, not the speaker — Stopped capture (pid …), the Readable: / Structured: paths, and where the dictation was saved: either handover: lines from the project's copilot.dictationHandoverCommand, or — when the project configured none — Saved to the repo: docs/inputs/copilot/… and a To commit: line (the copy is never committed for you), or a Not saved to the repo (…) line, which means the recording was kept back on purpose (a public-host remote stands the built-in target down) — report it as kept, do not retry. Ignore them; everything else is the message.

Treat that text as the user's message — act on it. Do NOT echo it back. Respond in the language the user dictated in. If no text was captured, say: "Dictation stopped, no text captured."