Skip to content

Discover commands, skills, and collaborators

Proqi 0.11.0

Proqi has two discovery systems. Commands finds Proqi actions. The inline invocation picker finds local agent skills, slash commands, and recognized collaborators while you edit a thought.

Search all Proqi actions

Press : in Board mode. Commands initially shows a short Relevant now set, then searches every searchable Commands entry as you type. Ranking is deterministic and Unicode-aware.

Unavailable commands are filtered by the current context and state. For example, Extract selection as new thought requires an editor selection, and Retry Screenshot Capture requires a retained failed capture.

Use arrows to move, page keys or Alt plus vertical arrows to jump, Enter or a click to run, and Escape to close. Printable j and k remain query text in Commands. The Commands query has its own transient undo history and never falls through to hidden Board history.

The Commands reference lists every action.

Use contextual Help

Press ?. Help is projected from the same action registry as Commands, keyboard dispatch, footer controls, and diagnostics. It shows only the controls relevant to the active Board, editor, or recovery context and uses effective configured bindings.

Complete a local invocation

While editing, type a supported sigil followed by part of a name:

  • $ for discovered skills;
  • / for discovered command or prompt definitions;
  • supported @ forms for recognized collaborators.

Exact and prefix matches rank first. Contiguous and separator-aware fuzzy matches follow, so a compact ordered abbreviation can find a longer hyphenated name. Sigils are hard namespaces: dollar, slash, and at results never mix.

Use arrows or Primary+P and Primary+N to navigate. Press Enter or Tab to insert the selected invocation, or Escape to close. Results beyond the visible viewport remain reachable by keyboard and mouse.

When a compatible adjacent Codex or Claude Code agent is verified, Proqi also offers a checked-in set of shared built-in commands at the first byte of the thought. The set is /btw, /clear, /compact, /diff, /fast, /goal, /hooks, /mcp, /model, /new, /permissions, /plan, /rename, /resume, /review, /skills, /status, /theme, and /usage. Leading whitespace, later lines, inline prose, extended names, and fenced code do not offer or highlight these built-ins. A valid prefix at byte zero can open completion; only a complete token is highlighted. If a discovered command has the same token, it retains this document-start-only placement. Choosing a row inserts text only. The receiving harness decides what the command does, and actual availability may depend on its version, account, platform, and feature configuration.

Refresh discoveries

Commands exposes separate refresh actions for invocations and adjacent agents. Use them after local definitions, plugins, packages, or Herdr topology change. Refresh never edits the thought by itself.

Understand collaborator mentions

Inside Herdr, the picker can discover recognized live coding agents on the same server. Selecting one inserts an inert collaborator location and renders it as a compact mention. It does not focus, message, or otherwise control that agent.

Actual prompt delivery remains a separate deliberate action. See Deliver prompts to agents.