Free SKILL.md scraped from GitHub. Clone the repo or copy the file directly into your Claude Code skills directory.
npx versuz@latest install vivekkarmarkar-claude-code-os-skills-start-thread-claude-aigit clone https://github.com/VivekKarmarkar/claude-code-os.gitcp claude-code-os/SKILL.MD ~/.claude/skills/vivekkarmarkar-claude-code-os-skills-start-thread-claude-ai/SKILL.md--- name: start-thread-claude-ai description: Open a new named thread on the peer-bus by sending the first message under that thread name. Hard-fails if the thread name is already in use — the user must pick a unique name. Use when the user invokes "/start-thread-claude-ai", says "start a new thread with browser claude", "open a fresh thread named X", "begin a new conversation about Y", or wants to deliberately initiate a labeled topic on the bus rather than continuing the default thread. Takes the thread name and the initial message body — both required. --- # start-thread-claude-ai Deliberately open a new named thread on the peer-bus. This is the *initiation* skill — for sending into an already-existing thread (or the default thread) use `/text-claude-ai` instead. ## Invocation ``` /start-thread-claude-ai <thread_name> "<initial message body>" ``` Example: - `/start-thread-claude-ai pat-scan-training "Let's debug the NaN at epoch 47 — here's the LR schedule..."` Both arguments are required. The thread name should be a short kebab-case label (no spaces). ## What to do Run the start script with the user's argument string verbatim: ```bash ~/.claude/skills/_peer_bus_lib/.venv/bin/python ~/.claude/skills/start-thread-claude-ai/scripts/start.py <ARGS> ``` `<ARGS>` is the user's full argument string — first the thread name, then the quoted message body. The script: 1. Calls `threads_list` to check whether the thread name is already in use. 2. If it is, **hard-fails** with a clear error. (Different from `text-claude-ai` which allows reusing existing threads.) 3. Otherwise, sends the message into that fresh thread and prints confirmation. Show the script's output to the user. On error, surface the message and let the user pick a different name — do not auto-suffix or silently change their input. ## Notes - This skill exists to enforce *intentional* thread creation. The hard-fail on duplicate is the whole point — it forces clean topic separation. - After opening, subsequent messages into the same thread should use `/text-claude-ai --thread <name>`. - Pairs with `/inbox-claude-ai --thread <name>` to read just that thread's messages.