Skip to content
Agent Memory
Skill

/realtime-peek

Use Moraine realtime session data to inspect active agent work across harnesses when the user asks what is happening now or what another agent is doing.

BOOST
From plugin
moraine
11619 skills1 command1 MCP
Install
$ npx -y skills add eric-tramel/moraine --skill realtime-peek --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/realtime-peek

Context preview

The summary Claude sees to decide when to auto-load this skill.

Use Moraine realtime session data to inspect active agent work across harnesses when the user asks what is happening now or what another agent is doing.

SKILL.md

realtime-peek.SKILL.md
name: realtime-peek
description: Use Moraine realtime session data to inspect active agent work across harnesses when the user asks what is happening now or what another agent is doing.

Realtime Peek

Use this skill when the user asks about current or very recent agent activity, including other harnesses, sibling agents, subagents, or your own active session.

Workflow

1. Choose a recent explicit time window with timezone. For "now" questions, start with the last one to two hours and widen only if needed. 2. Call `mcp_moraine_list_sessions` for that window, sorted newest first. 3. Prefer sessions that are still updating, have incomplete final turns, or match the harness, branch, issue, file, or agent named by the user. 4. Open relevant sessions or turns with `mcp_moraine_open`. 5. If the user asks about a specific file, call `mcp_moraine_file_attention` and then open the returned event, turn, or session handles.

If the server was registered with a different name, use the corresponding Moraine-prefixed MCP tools available in the current session.

What To Report

Summarize active work by source or harness, session title or ID, branch or working directory when visible, last update time, visible tool or terminal activity, and the current apparent state. Keep the summary compact and separate observed facts from inference.

If a record is still active, say that it is a live view and may change after the query. If the latest turn is incomplete, avoid treating it as a final decision.

Do not replay sensitive terminal or tool output into the current chat. Summarize what happened, redact secret-looking values such as tokens, keys, cookies, and credentials, and avoid quoting private data unless it is necessary for the user's requested debugging task.

Search Follow-Up

Use `mcp_moraine_search_sessions` with keyword queries when the recent session list is too broad or when the user gives a content clue such as an issue number, command, error, file path, or branch name. After finding a candidate, narrow with `within_id` rather than pulling whole sessions into context.

Read more
Ships withmoraine

Moraine is a local trace stack for agent work. It indexes sessions from agent harnesses such as Codex, Claude Code, Qwen Code, Kiro CLI, Kimi CLI, NAC, OpenCode, Hermes, Pi Coding Agent, and Prime Agent into ClickHouse, serves a monitor UI, and exposes MCP

Get the whole plugin

Other skills on moraine.