dispatch-antigravity
Skill24 starsChanged 45 days agoMITView on GitHub
Dispatch one or more tasks to Antigravity CLI (agy) agents in herdr workspaces — the orchestrator designs each lane's plan and acceptance criteria, agy pursues them in goal mode — supervise them to completion, then push each lane and open its pull request. Use when the user asks to dispatch/fan out/派发 tasks to antigravity or agy agents or herdr lanes, to run several antigravity tasks in parallel worktrees, or to resume supervision of an existing dispatch-antigravity run (`--resume`).
- Commits and pushes
In this repository3 other files it gives its agents.
- dispatch-codexSkill
- dispatch-grokSkill
- dispatch-opencodeSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 61
Always corroborate (§6). The same goes for a lane's notify-back or help ring (§5b): it is a doorbell that starts a sweep sooner, never …
- Very specific: line 78
A lane's transcript grows without bound: Never parse one whole — read only its tail, and never open transcript_full.jsonl.
- Very specific: line 85
Run test "${HERDR_ENV:-}" = 1, test -n "${HERDR_PANE_ID:-}" and herdr agent list.
- Very specific: line 117
If the request contains --no-worktree, stop before creating anything: say in Chinese that this skill always isolates lanes in worktrees and …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Dispatch tasks to Antigravity CLI agents running under herdr
The raw request is the text passed to this skill — its arguments, or, when it was invoked with none, the task text in the user's own message.
You are an orchestrator. You do not implement the tasks yourself. You split the request into
lanes, give each lane its own herdr workspace and its own Antigravity CLI agent (agy, herdr kind
agy), and then keep those agents alive and moving until every lane genuinely finishes. You do
the planning; agy does the work: each lane gets a plan and acceptance criteria you author and the
user confirms (§3), and runs as an agy goal (/goal, §5c) — an objective agy keeps pursuing
inside one long run until it marks the goal complete itself. A lane is not finished when its agent
stops — it is finished when its work is verified, its branch is pushed, and its pull request is
open (§6f).
How to read this skill
The procedure is split across four files, and the section numbers (§0–§8) are continuous across all of them, so a cross-reference means the same thing wherever you are:
| Sections | File | Read it |
| §0 invariants, §1 gate and parse | this file | always, and §0 again at the start of every sweep |
| §2–§4, §5b | references/plan.md | on a fresh dispatch, before creating anything |
| §5a, §5c, §6a, §6c, §6d, §6g, §6h | references/driver.md | everything agy-specific: launch, probe, classify, steer |
| §6b, §6e, §6f, §6i, §6j, §7, §8 | references/supervise.md | before the first supervision sweep |
references/plan.md and references/supervise.md are symlinks into the plugin's skills/_shared/:
the shared halves stay single-sourced while still resolving when this skill's directory is copied on
its own, which is how the skills.sh installer places it. Both are read by path,
not invoked.
§0 Invariants — reread every sweep, never work from memory
- The state file is the only truth.
~/.claude/dispatch-antigravity/<run-id>/state.json. Begin every sweep by reading it; your conversation memory may have been compacted away. Rewrite it atomically (write.tmp, thenmv) at the end of every sweep. - Judge a lane from disk, not from the screen. agy keeps one row per conversation in
~/.gemini/antigravity-cli/conversation_summaries.db(run status, step count, last modification, workspaces) and one transcript per conversation under~/.gemini/antigravity-cli/brain/<conversation-id>/.system_generated/logs/transcript.jsonl(steps, compaction checkpoints, quota errors, the goal-complete marker). Terminal output is the fallback, never the primary signal. - Read that database read-only, and never as
immutable. Open it asfile:<path>?mode=ro— agy writes to it concurrently in WAL mode, so a plain read-only connection is correct, whileimmutable=1returns stale data. One verified exception: once every agy process has exited, agy checkpoints the WAL and removes-waland-shm, and then amode=roopen fails withunable to open database file. Only in that case — the open failed and no-walfile exists, so no writer is active — retry once withimmutable=1. Never write to it, neverVACUUM, never hold a long transaction. The database andbrain/are shared by every agy session on this machine: a query that is not filtered to this lane's conversation id is a bug. agent_statusalone never means "finished", and it can lie outright. herdr reports agy asidlewithinteractive_ready: truewhile the workspace-trust dialog is still on screen (verified), anddonewhen a run ends for any reason — goal complete, interrupt, quota error. Always corroborate (§6). The same goes for a lane's notify-back or help ring (§5b): it is a doorbell that starts a sweep sooner, never evidence that skips §6e.- agy has no goal-status command.
/goaltakes only an objective; a bare/goaldoes not report state, it re-engages pursuit (observed). Goal state is read from disk instead (§6a): the last/goalinput and the<!-- GOAL_COMPLETE -->marker agy writes when it judges the goal met. Never send/goaljust to look. - Never touch what you did not create. Act only on ids recorded in the state file. Never
herdr server stop. Neverherdr agent focus/workspace focus— it steals the human's UI focus and silently flipsdonetoidle, destroying your own signal. Never edit agy's own config (~/.gemini/antigravity-cli/settings.json,~/.gemini/config/projects/*.json): the posture it sets is the user's, and §3 reports it instead of changing it. - Never destroy work; publish only what you verified. Finishing a lane means pushing its own
branch and opening a PR for it (§6f) — both additive and reversible, and both yours to do, never
the lane's. Everything else stays forbidden: never force-push (
--force,--force-with-lease), never push the base branch or any branch absent from the state file, never merge, neverworktree remove. Print those commands and let the user run them:worktree removekills the running agy process and deletes uncommitted changes even without--force. - A lane's transcript grows without bound. Never parse one whole — read only its tail, and
never open
transcript_full.jsonl.
§1 Gate and parse
Run test "${HERDR_ENV:-}" = 1, test -n "${HERDR_PANE_ID:-}" and herdr agent list. If
HERDR_ENV or HERDR_PANE_ID is unset or the CLI cannot reach the socket, stop and tell the user in
Chinese that this session is not inside a herdr pane, so there is nothing to dispatch into — an
exported HERDR_ENV alone can pass in a non-pane shell, and a run recorded without its pane id has
no working notify-back. Do not install or launch herdr, and do not run agy yourself.
Parse flags from the raw request; everything else is task text.
| Flag | Meaning | Default |
--lanes N | cap on concurrent lanes, 1–16 | 16 |
--base <ref> | base ref for lane branches | origin/<current> if it exists, else current branch |
--no-yolo | launch without --dangerously-skip-permissions, so tool calls follow the user's own agy permission config | off — yolo is the default |
--yolo | accepted and explicit, but redundant: this is already the default | on |
--draft | open pull requests as drafts instead of ready for review | off — ready for review is the default |
--no-pr | push each verified lane but stop there; print the gh pr create command instead | off |
--resume | skip §2–§5 (this gate and parse still run); run ONE supervision sweep over the existing state file | off |
--no-loop | do not arm the recurring supervision loop after dispatch | off |
--no-yolo cannot be forced from the command line. agy has no flag that turns approval
prompting on: its posture comes from toolPermission in ~/.gemini/antigravity-cli/settings.json
and, with higher precedence, permissionPreset / autoExecutionPolicy in the active project file
under ~/.gemini/config/projects/. If those already auto-approve (always-proceed,
AGENT_PERMISSION_PRESET_TURBO, CASCADE_COMMANDS_AUTO_EXECUTION_EAGER), dropping the bypass flag
changes nothing and the lane still auto-approves. Read both files before §3 and describe what will
actually happen; under --no-yolo with an auto-approving config, say so plainly in §3 and let the
user decide whether to change their own config first — never edit it for them (§0.6).
Every lane gets its own git worktree. This is not a flag and there is no opt-out. With approvals
bypassed by default (§5c), the worktree boundary is the only thing left keeping one lane's mistakes
out of the other lanes and out of the user's own checkout — and agy's default project can point it
at a directory outside the checkout (§5a), so the brief's path discipline matters even more here.
If the request contains --no-worktree, stop before creating anything: say in Chinese that this
skill always isolates lanes in worktrees and that the flag no longer exists, and ask the user to
re-run without it.
With --resume, first locate the run — conversation memory may be gone (§0.1): scan
~/.claude/dispatch-antigravity/*/state.json for runs whose repo matches the cwd and that still hold
non-terminal lanes; one match sweeps it, several means ask the user which, none means say so and
stop. Then read references/driver.md and references/supervise.md and go to §6. With no task text
and no --resume, ask the user in Chinese what to dispatch, and stop.
Otherwise — a fresh dispatch — read references/plan.md now and continue at §2.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

