Skip to content
Operations
Skill

/tma1-peer

List recent sessions on this project by agent — peer agents (Codex, OpenClaw, Copilot CLI) or your own past sessions. Invoke when the user wants you to read another agent's review feedback, see what someone else tried, act on cross-agent context, or recall your own earlier work

BOOST
From plugin
tma1
1204 skills2 commands
Install
$ npx -y skills add tma1-ai/tma1 --skill tma1-peer --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/tma1-peer

Context preview

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

List recent sessions on this project by agent — peer agents (Codex, OpenClaw, Copilot CLI) or your own past sessions. Invoke when the user wants you to read another agent's review feedback, see what someone else tried, act on cross-agent context, or recall your own earlier work

SKILL.md

tma1-peer.SKILL.md
name: tma1-peer
description: "List recent sessions on this project by agent — peer agents (Codex, OpenClaw, Copilot CLI) or your own past sessions. Invoke when the user wants you to read another agent's review feedback, see what someone else tried, act on cross-agent context, or recall your own earlier work here. Trigger phrases: \"what did codex do\", \"what did openclaw do\", \"what did copilot do\", \"peer sessions\", \"cross-agent context\", \"what did I do earlier\", \"我之前在这个项目做了什么\", \"/tma1-peer\"."
argument-hint: "[agent] [--full|--summary] [--limit N] [--since 2h]"
allowed-tools: ["mcp__tma1__get_peer_sessions", "mcp__tma1__get_session_transcript"]

TMA1 Peer-Agent Lens

Lists recent sessions on the current project, newest first. Default is peer agents — the ones that aren't you — because the usual reason to call this is "another agent reviewed something, act on it without me copy-pasting between terminals". Naming yourself is also allowed.

Syntax

/tma1-peer                          # all peers, latest session each
/tma1-peer codex                    # codex only
/tma1-peer codex --full             # codex, with full conversation content
/tma1-peer self                     # your own recent sessions on this project
/tma1-peer codex --limit 3          # codex, latest 3 sessions
/tma1-peer --summary --limit 5      # all peers, headers only, 5 each
/tma1-peer codex --since 6h         # widen the time window

Agent names: `codex`, `openclaw`, `copilot` (= `copilot_cli`), `claude` (= `claude_code`), `self` (= you), `all` / omitted (= every peer but you). Anything else: reply `unknown agent "<X>"; available: codex, openclaw, copilot, claude, self, all` and STOP — do not call the tool with an unrecognised name.

Flags

| Flag | Maps to | Default | | --- | --- | --- | | `--summary` | `detail: "summary"` — header, tools, files; no messages | | | (none) | `detail: "preview"` — 3 recent messages per session | default | | `--full` | `detail: "full"` — 40 recent messages per session | | | `--limit N` | `limit` — sessions per agent, 1-5 | 1 | | `--since 2h` / `--since 90m` | `since_min` in minutes | 1440 (24h) |

Legacy positional form still works: `/tma1-peer codex 3 30` means agent `codex`, `limit: 3`, `message_limit: 30`. A bare leading integer is the limit, not an agent (`/tma1-peer 3` → all peers, `limit: 3`). Prefer the flags above when writing new invocations.

Call the tool

`mcp__tma1__get_peer_sessions` with the parsed `agent_source`, `detail`, `limit`, `since_min`. Pass every value the user actually supplied — do not silently fall back to the defaults.

Use the response

  • `sessions[]` — `session_id`, `agent_source`, `last_activity_ago`,

`duration_minutes`, `tool_call_count`, `messages`, `recent_tool_names`, `files_touched`, `cwd`. On a `self` lookup, `is_latest_for_cwd` marks the session that is almost certainly the one you're in right now — don't report your own current work back to the user as news.

  • `most_recent_session` — freshest session's agent + age. Lead with it so

the user knows immediately whether the peer work is current.

  • `partial_failures` — `agent → error`. Present only when the all-peers

fan-out had a failure. **Read it before calling empty results silence.**

  • `note` — present only when `sessions` is empty and nothing failed.

Rules:

  • **Quote what the peer actually said.** Do not paraphrase. Reading the

peer's exact words is the point of this command.

  • Messages are capped per `detail`; a cut message carries `truncated: true`.

Need the rest? `mcp__tma1__get_session_transcript` with that `session_id`.

  • Empty `sessions`, no `partial_failures` → say `no recent <agent> sessions

on this project in the window` and stop. Don't fabricate.

  • Empty `sessions` with `partial_failures` → name the peers that failed and

quote the error instead of asserting silence.

Related

  • Looking for a session by keyword rather than by agent? Use `/tma1-search`.
  • Already have a `session_id`? Go straight to `get_session_transcript`.

Examples

User: `/tma1-peer codex --full` You: (call with `agent_source: "codex", detail: "full", limit: 1`) You: "Codex reviewed `auth.go` 12 min ago and left three concrete issues: 1. ... 2. ... 3. ... Want me to address all three or pick one?"

User: `/tma1-peer self --limit 3` You: (call with `agent_source: "self", limit: 3`) You: "Three of your sessions here today. The current one aside, the 14:20 session landed the retry fix and the 11:05 one was the failed migration attempt."

Read more
Ships withtma1

pre-release, and without it the badge renders "no releases found". Leave sort at the default (date) — semver sort compares the pre-release identifier lexically, which orders alpha9 above alpha14. --> A monolith for your agent's loop.

Get the whole plugin
Stats
121
Stars
14
Forks
Active
Maintenance
Go
Language
Apache-2.0
License
2d ago
Last commit
7mo ago
Created
3h ago
Added

Repo: tma1-ai/tma1

Other skills on tma1.

tma1-search
Skill

tma1-search

Search past agent sessions on this project by keyword, and read back any session's conversation. Invoke when the answer lives in…

@tma1-ai@tma1-aiInstall Skill
tma1-setup
Skill

tma1-setup

Install and configure TMA1 local observability (Claude Code, Codex, GitHub Copilot CLI, OpenClaw, any OTel SDK). Use when the…

@tma1-ai@tma1-aiInstall Skill
tma1
Skill

tma1

Query TMA1 observability data (Claude Code, Codex, GitHub Copilot CLI, OpenClaw, any OTel SDK). Use when the user asks: how much…

@tma1-ai@tma1-aiInstall Skill