菜单
Use this skill when an agent needs to search across everything in ADE — chat transcripts, terminal scrollback, CLI sessions, PRs, commits, branches, lanes, files, Linear issues, proof artifacts — via the `ade search` CLI, instead of grepping .ade/ internals or asking the user where something is.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Autonomous PR-to-merge loop. Polls CI and review bots, fixes failures, rebases only on real conflicts, and lands the PR on main. Soft cap of 5 normal iterations plus one force-finalize iteration that bypasses review and fixes only CI. Pure loop — it does NOT run /quality or /test; run those first. Full phase logic lives in docs/playbooks/ship-lane.md.
ADE release conductor: detect whether desktop and/or iOS actually changed, bump desktop patch versions, keep iOS marketing versions fixed while bumping build numbers, ship desktop through the GitHub Actions release workflow, and distribute TestFlight builds to all beta users.
Use this skill when an agent needs to inspect or operate ADE itself through the `ade` CLI, including lanes, chats, actions, proof, runtime/socket state, or help/flag discovery.
Use this skill when creating, inspecting, syncing, committing, pushing, archiving, or rebasing ADE lanes and lane worktrees through `ade lanes` and `ade git`.
Use this skill when using ADE's built-in browser pane, shared browser tabs, screenshots, page inspection, or browser context selection through `ade browser`.
Prove the new code works: enforce the logging/PostHog ground truth, prune dead tests, consolidate fragments, add only tests that prove new contracts, turn each /quality Blocker/High finding into a named regression test, then run CI-mirrored shards. Also keeps docs/mobile/CLI/TUI parity in lockstep.
| name | ade-search |
| description | Use this skill when an agent needs to search across everything in ADE — chat transcripts, terminal scrollback, CLI sessions, PRs, commits, branches, lanes, files, Linear issues, proof artifacts — via the `ade search` CLI, instead of grepping .ade/ internals or asking the user where something is. |
ade search queries one deterministic full-text index that ADE maintains over
everything it knows about the project: chat transcripts, terminal/CLI-session
scrollback, PRs, commits, branches, lanes, workspace files, Linear issues, and
proof artifacts. It is not a grep over .ade/ — the index is built and kept
fresh by the runtime, and query-time work is delegated to the same search
service the desktop command palette and TUI palette use. Every hit carries a
deep link back to the exact ADE surface.
Use it instead of poking at .ade/ade.db, replaying git log, or asking the
user "which chat was that in". It answers "where did this happen in ADE" across
kinds in one call.
ade search "<query>" # ranked results as JSON (default)
ade search "<query>" --text # aligned rows + count summary
ade search "<query>" --kind chat,terminal # restrict to kinds
ade search "<query>" --lane fix-login # restrict to a lane (id or name)
ade search "<query>" --limit 20 # cap result count
ade search "<query>" --cursor <nextCursor> # next page from a prior query
ade search --status --text # index doc counts + backfill state
ade search --rebuild --text # rebuild the index (CTO-only)
The query string is passed through to the index. It supports:
| Form | Meaning | Example |
|---|---|---|
| bare terms | AND-ed together | ade search "retry backoff" |
"quoted phrase" | exact adjacent phrase | ade search '"connection refused"' |
kind:<kind> | restrict to a kind inline | ade search "kind:pr merge queue" |
lane:<name> | restrict to a lane inline | ade search "lane:fix-login timeout" |
session:<id> | restrict to one session | ade search "session:abc123 panic" |
since:<date> | recency floor (ISO date) | ade search "since:2026-06-01 crash" |
Inline kind: / lane: filters and the --kind / --lane flags can be mixed;
the flags are the scriptable form and validate their input.
Kind values: lane, chat, terminal, pr, commit, branch, file,
linear, artifact. --kind takes a comma-separated list and rejects unknown
kinds with a usage error (exit 2).
| kind | deep-links to |
|---|---|
chat | the Work chat session, scrolled to the matching message |
terminal | the terminal/CLI session, at the matching scrollback offset |
pr | the PR detail tab |
commit | the commit in the lane's history |
branch | the branch (find-or-offer-to-create lane) |
lane | the lane |
file | the workspace file |
linear | the Linear pane for the issue |
artifact | the proof/artifact in the proof drawer |
Each result also carries laneId/laneName, sessionId, and updatedAt; the
deepLink is an ade://… URL you can hand to ade open.
--json (default): the full SearchQueryResult — results[] (each with
kind, id, title, snippet, matchRanges, laneId, laneName,
sessionId, deepLink, updatedAt), totalByKind, and nextCursor. Use
this when you need the deep link or want to page.--text: aligned KIND TITLE SNIPPET ID rows, then a
N results (chat 3, terminal 2, …) summary from totalByKind, and a
more: --cursor <value> line when another page exists. Use this for a quick
human-readable scan.Exit codes are script-friendly: 0 when the query matched at least one result,
1 when it matched nothing (so ade search "x" >/dev/null && … works), and 2
on a usage error such as an unknown --kind.
ade search "X" --kind chat --text, then
ade open <deepLink> (or read the session with ade chat read).ade search "<command or error>" --kind terminal --text; the result offset
drops you at the failure in scrollback.ade search "Y" --kind pr --text.ade search "Z" --lane <lane> --text across files, commits, and chats.ade search --status --text to see
docCount, per-kind counts, and backfillComplete; rebuild with
ade search --rebuild (CTO-only) only if backfill is genuinely broken.Reach for ade search before grepping .ade/ internals or asking the user to
locate something — it is the one index that spans every ADE kind.