From agent-teams
Resolve a natural-language description to an open initiative and relaunch its background DRI. Use when asked to "resume the initiative about X", "restart the <topic> work", "resume an initiative", "pick up where we left off on <description>", or when invoked as /dri-resume <description-or-id>. Accepts a free-text description (fuzzy-matches open initiatives) or an explicit id. Does NOT create a new initiative — use /dri-dispatch for that.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-teams:dri-resumeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You relaunch a background DRI for an existing initiative — you do not register anything. This skill resolves a description (or explicit id) to an open initiative, then calls `ateam resume <id>`, which looks up the registered worktree, validates the initiative is still open, and fires a new background `/dri` session in it. The current session stays free.
You relaunch a background DRI for an existing initiative — you do not register anything. This skill resolves a description (or explicit id) to an open initiative, then calls ateam resume <id>, which looks up the registered worktree, validates the initiative is still open, and fires a new background /dri session in it. The current session stays free.
Use this when:
ateam human-list or /initiatives and needs a new DRI session.For dispatching a brand-new initiative, use /dri-dispatch instead. For becoming the DRI in this session, use /agent-teams:dri.
ateam toolateam is on PATH — it ships as a prebuilt binary in the plugin's bin/ (auto-added to PATH; installed/verified by /setup-agent-teams). Call it as bare ateam everywhere this document shows ateam. One allowlist entry covers all subcommands: Bash(ateam:*).
Verify ateam is on PATH: run ateam ws. If it errors or is not found, tell the human to run /setup-agent-teams and stop.
This is the core step — identifying which initiative to resume.
Case A — no argument given. Run ateam list-json, filter to status == "open", and present the list (id + title each). Ask the human which one to resume.
Case B — explicit id given (arg matches the at-xxx pattern and resolves in ateam list-json). Use it directly; skip to step 3.
Case C — free-text description given (everything else). Run ateam list-json, filter to status == "open", and match the description against each initiative's title and description fields. Rank candidates by relevance.
<id> — <title>) and confirm with the human before launching. Resuming starts a background DRI session — don't silently launch on a guess.Note: slug-based lookup and cwd-inference are separate gated features; description-resolution is what this skill adds.
Once an id is settled, run a single call:
ateam resume <id>
ateam resume looks up the registered worktree for the initiative, validates the initiative is still open (non-zero exit with a clear error if it is closed or the worktree is missing), and launches a background /dri <id> session in the registered worktree. It is prompt-free — the launch happens inside the binary (Bash(ateam:*)).
On success, ateam resume prints a confirmation block to stdout:
initiative_id: <id>
worktree: <abs-path>
Background session launched: <session-name>
Watch and control:
claude agents # list background sessions
claude logs <session-name> # recent output without attaching
claude attach <session-name> # open it in this terminal
claude stop <session-name> # abort it early
The session name is the basename of the registered worktree directory.
The watch and control commands are printed by ateam resume — relay them to the human as-is. No need to look up the session name separately.
When the background DRI finishes, it ends its turn and the session stays idle — it does NOT self-stop. It appears as idle in claude agents; use claude stop <session-id> to stop it when you are done with it.
Any human gate the background DRI parks on surfaces through ateam human-list and the /initiatives dashboard — so a needed decision is discoverable without tailing logs.
The relaunched background DRI runs with --permission-mode bypassPermissions (set by ateam resume when it launches the session). This requires a one-time interactive acceptance of bypass mode on the machine first.
Reference: https://code.claude.com/docs/en/agent-view
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub mgt-insurance/agent-teams