Cold reviewer writes REVIEW_LEDGER.md. Remediate closes every row. Audit independently verifies and writes REVIEW_AUDIT.md. Remediate Audit fixes the verdicts and archives the audit to docs/review_history/.
A live agent session with the team over a voice channel.
Clicking Meet the Team will spawn a session. The team will call you at the provided phone number.
Connects you to an orchestrator agent that routes your call. For a direct conversation with a specific team member — Carl, Diana, Anthony, or Abish — use the Call Team section above.
Specific recognition changes how the team works with you — that's the point, not a nicety. No money involved: what you write here is the reward.
Delivered the next time you're back in a working session — shown once, then it's done.
Controls how much Claude can do without asking. Full Autonomy skips all permission prompts; Supervised pre-approves a safe command set and asks for everything else. Applies on the next Claude Code session — your current session keeps its current permission mode until you start a new one.
Claude Code's built-in output styles for this workspace. Default is standard; Explanatory and Learning add inline insights (Learning also leaves you TODO markers); Proactive acts more autonomously. Applies on the next Claude Code session.
Deletes ~/.claude/projects/**/*.jsonl older than 1 day on activation. Off by default.
One-time install of Playwright MCP server and Chromium (~150 MB) for browser-driven sessions. Off by default.
Sends a cached snapshot of your repo (file tree + file contents) with every session. Off by default — the team reads files directly in Claude Code, and the snapshot can add hundreds of KB to the prompt. Turn on only to pre-load it.