tomevault-io bullshit — Cross-LLM Fact Checker
Dispatches your conversation to another LLM (codex, copilot, gemini, aider) for independent fact-checking.
Paths
All scripts live in scripts/ relative to this skill's base directory.
Use the base directory shown when this skill loaded to construct absolute paths.
In commands below, <base> = that directory.
Commands
| Command | Action |
|---|---|
/bullshit |
Dispatch fact-check to preferred CLI |
/bullshit setup |
Configure preferred CLI and options |
/bullshit sync |
Re-scan available CLIs (bust 24hr cache) |
/bullshit read |
Manually read latest result (fallback) |
Dispatch (default)
Run send.sh in background. The session ID for this session is: ${CLAUDE_SESSION_ID}
bash "<base>/scripts/send.sh" ${CLAUDE_SESSION_ID} [N]
Second arg = number of conversation messages to review (default: 10). This counts user/assistant turns only — tool calls are excluded. Use more for thorough checks, fewer for quick ones.
Run this with run_in_background: true. Tell user: "Dispatched to {cli}. I'll get the results when it's done — keep working."
When background task completes, you receive the fact-check output. Present findings. Think critically — the other LLM can be wrong too. Verify before accepting or dismissing.
Setup
Run bash "<base>/scripts/detect-clis.sh" to show available CLIs. Ask user to pick preferred CLI and context size. Write to ~/.config/bullshit/config.json:
{"preferred_cli": "codex", "context_messages": 10, "max_chars": 50000, "timeout_seconds": 300}
Sync
bash "<base>/scripts/detect-clis.sh" --bust-cache
Read (fallback)
Only if background delivery failed. Find latest: ls -t /tmp/bullshit-*.json | head -1
Source: rgr4y/bullshit — distributed by TomeVault.