* feat(agents): mobile_ui agent tool for Android UI control (PR 3/3)
Add a dedicated model-facing tool that drives another Android app through the
PR2 mobile.ui.observe/act node commands, completing the vertical slice
(agent -> tool -> node.invoke -> AccessibilityService).
- Mirrors the desktop computer tool's safety model: owner-only + HTTP-denied
(dangerous-tools), raw node.invoke of mobile.ui.* redirected to this tool so
the generic nodes tool cannot bypass it, run/tool-call idempotency, and the
phone-arm workflow (mobile.ui.* must be explicitly armed).
- One call is observe or one act; every act automatically re-observes for
postcondition verification and preserves the landed outcome if re-observe fails.
- Fail-closed confirmation: all state-changing acts (activate, set_text, tap,
swipe) require confirmed=true after the model reviews the proposed effect;
observe, scroll, wait, and navigation global_actions do not. The keyword list
only enriches the confirmation message and is never the sole gate (a11y labels
are localized/iconographic/coordinate-blind).
- Node selection resolves an explicit id against the full device set first
(case-insensitively) and rejects an ineligible or ambiguous match, so an
explicit selection can never be silently redirected to the wrong phone.
- All observed UI text is treated as untrusted; the tool description forbids
following instructions found in app UI.
Additive; no protocol bump. mobile-ui-tool + registration + policy tests pass;
core tsgo and lint verified on Testbox. On-device drive is exercised via the
PR1/PR2 emulator proof; this PR is the agent-side tool + tests.
Follow-up: computer-tool.ts has the same latent node-resolution ordering bug
(explicit id searched only among eligible nodes); tracked separately.
* fix(agents,phone-control): migrate mobile_ui arming hints/tests to gateway.nodes.commands.allow
Match main's node-command arming rename (gateway.nodes.allowCommands ->
gateway.nodes.commands.allow / commands.deny): update the mobile_ui tool arm-hint
matchers to the gateway's current rejection strings and fix the phone-control /
tool tests to the current config shape. Production write path was already correct.
* fix(agents): register mobile_ui in owner-only denylist; trim phone-control under max-lines
- tool-resolution.exclude.test.ts: mobile_ui joins the owner-only core
tools, so add it to the expected plugin/inherited denylists.
- phone-control: derive ArmGroup from a single ARM_GROUPS const and
collapse parseGroup's ||-chain, dropping index.ts back under 700 lines.
* fix(agents): centralize mobile_ui global-action names; regenerate tool-display snapshot
- Extract GLOBAL_ACTION_NAMES const so the schema, action type, and
validator share one source. This also removes the bare `name: "back"`
type-annotation literal that scripts/tool-display.ts's name-regex was
misreading as a phantom runtime tool.
- Regenerate apps/.../tool-display.json to add the mobile_ui display entry
(was missing from the Swift snapshot).
* refactor(channels): trim deprecated setup-input fields and legacy promotion keys to published readers
* ci: retrigger after dropped ready event
* fix(channels): update setup-input guardrails and dead code for the trimmed tier
Media-facts program PR 5 — the planned additive bridge. Structured
facts travel alongside existing prompt text through prompt prelude,
GetReplyOptions, FollowupRun, queue collect/defer/retry, active
steering, embedded/CLI runs, Gateway offloads (offloadedRefs no longer
discarded; AgentContentPhaseResult carries media), Tlon, and late-media
projection — with prompt bytes golden-equal everywhere. Facts follow
the same adoption/idempotency/late-append lifecycle as prompt text.
Intentionally +175 prod LOC per the audit; the facts-first
hydration/prune deletion PR consumes this carrier next.
* fix(memory): keep canonical rows when same-file legacy memory tables diverge
* fix(memory): keep canonical-owned sources' chunk sets coherent during legacy import
* fix(memory): import legacy chunks when canonical source has no chunks
The same-file legacy migration excluded a legacy chunk from import whenever the
canonical index already had a source row for its (path, source). That stranded a
file whose canonical source was registered but had no chunks yet (indexing
interrupted before chunks were written, or embedding pending/failed): the legacy
chunks were its only searchable content, and the matching source hash stops sync
from re-indexing, so the file went silently unsearchable.
Re-key the chunk-coherence exclusion on canonical chunk ownership instead:
snapshot the (path, source) pairs that already have canonical chunks before the
import and skip legacy chunks only for those. A source with a canonical row but
no chunks now imports its legacy chunks. The snapshot is taken pre-insert because
the exclusion predicate reads the chunks table the import writes to.
Add regressions: legacy chunks import for a chunk-less canonical source while a
chunk-owning source still drops its stale legacy chunk; and restore abort
coverage for the meta and chunks copy assertions (previously only files was
exercised).
* fix(memory): harden same-file legacy conflict recovery
* fix(memory): rebuild ambiguous partial legacy sources
* fix(memory): reconcile migrated derived indexes
* fix(memory): close migrated index ownership gaps
* test(memory): align migration expectations
* test(tooling): match routed test order
* test(memory): exercise vector reload cleanup
* test(memory): prove real vector reload cleanup
* fix(memory): make migrated indexes converge
---------
Co-authored-by: Serhii Leniv <leniv.tech@gmail.com>
Co-authored-by: VACInc <3279061+VACInc@users.noreply.github.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(channels): restore transcript context after restart
Merge bounded active-branch session transcript turns at the shared prepared-turn seam so message channels retain assistant replies after restart or history eviction. Migrate Telegram's one-off merge while preserving exact projection and legacy dedupe behavior.\n\nCloses #112520. Slack case reported by Joe Tam (@joetam) in #102594.
* style(channels): avoid spread in transcript mapping
* refactor(telegram): drop obsolete transcript exports
Ensure doctor runtime tool schema checks and wiki CLI resolve agent-scoped vaults from canonical agents.entries while retaining the internal list projection fallback.
* refactor(channels)!: shrink ChannelSetupInput to a generic envelope with a deprecated compatibility tier
* fix(channels): keep ChannelSetupInput structurally assignable without an index signature
* docs: regenerate docs map
* fix(browser): tab creation steals window focus during agent automation
Agent-created tabs inherited CDP's foreground default: direct CDP
Target.createTarget omitted the background flag, and the extension
relay's createTab defaulted to active:true, so every agent tab open
activated the new tab (and, on the extension driver, focused the
window), interrupting whatever the human was doing in that browser.
Direct CDP tab creation now requests background:true (agent tab
ownership/selection is target-id based and never depended on
activation), and the extension relay defaults an omitted background
to true while preserving an explicit background:false, matching the
Codex/Claude-in-Chrome model the extension driver mirrors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor(browser): keep focus fix LOC-neutral
Preserve background tab creation while keeping the oversized CDP and relay modules within the current LOC ratchet.\n\nCodex-Session: 019f5e93-780a-7350-88f9-1986cdb64914
* fix(browser): honor explicit CDP focus requests
Keep background-by-default automation while treating Target.createTarget focus=true as an explicit foreground request in the extension relay.
Codex-Session: 019f5e93-780a-7350-88f9-1986cdb64914
* fix(browser): preserve explicit CDP focus semantics
Apply the background-by-default automation policy only when focus is omitted, preserving focus=false foreground-tab requests as well as focus=true.
Codex-Session: 019f5e93-780a-7350-88f9-1986cdb64914
* fix(browser): preserve create target window focus
Carry the resolved CDP focus intent through the extension relay and explicitly focus the containing Chrome window when requested.\n\nCodex-Session: 019f5e93-780a-7350-88f9-1986cdb64914
* style(browser): refresh relay import order
* test(secrets): use secure node exec fixtures
* test(doctor): secure exec secret fixture
* test(doctor): retain narrowed temp path
* test(secrets): secure remaining exec fixtures
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Persist qualified QQBot group and guild routes so session announcements return to the originating conversation without letting direct-message turns overwrite the shared route.
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Co-authored-by: lzyyzznl <lzyyzznl@users.noreply.github.com>
* fix(browser): retire durable tab rows whose browser never returns
Durable cleanup defers whenever ownership cannot be proven, so a browser
that never comes back at the same cdpUrl leaves its rows behind forever:
each sweep re-claims them, fails the identity lookup, warns, and defers
again. Nothing in the subsystem removes a row by age.
The `browser.session-tabs` namespace is opened with a 5000-row cap and
`reject-new`, so once those rows accumulate to the cap, tracking a new tab
throws PLUGIN_STATE_LIMIT_EXCEEDED. That propagates into the compensation
path in browser-tool-session-tabs.ts, which closes the tab the user just
opened and rethrows -- every `browser open` on that profile then opens a
tab, closes it again, and errors, with no self-healing path.
Bound the retry: when a close attempt reports the target unavailable and
the tab has been unused for longer than the retire window, drop the row
instead of deferring again. A browser returning after that long almost
always carries a fresh instance fingerprint, which retires the row through
the ownership-mismatch path anyway.
closeTrackedBrowserTabsForSessions now accepts `now` like the sweep does,
so lifecycle cleanup can be exercised on a coherent clock.
* refactor(browser): split session tab cleanup claim and test harness
check-lint failed on max-lines: session-tab-registry.ts was at 699 of its
700-line budget and the durable registry test at 982 of 1000, so the retire
branch and its regression test pushed both over.
Extract the cleanup claim bookkeeping (claim, ownership match, delete) into
session-tab-cleanup-claim.ts, and the durable registry test shapes into
session-tab-registry.sqlite.test-helpers.ts, matching the existing
*.test-helpers.ts convention in this directory. No behavior change.
* test(browser): protect unreachable retirement races
Co-authored-by: Yigtwxx <yigiterdogan023@gmail.com>
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(qa-lab): reject hex/exponent Telegram SUT uid env values
* test(qa-lab): add direct unit tests for parseSutId strict int parsing
Export parseSutId and add 15 direct unit tests covering hex/exponent/fraction/
empty/whitespace/zero/negative rejection and valid decimal acceptance across
UID, GID, and CLEANUP_TIMEOUT_MS keys. The 3 existing integration tests
already prove the full CLI chain rejects malformed UIDs before gateway startup.
Mutation check: reverting to permissive Number() parsing causes 6 tests to
fail (4 unit + 2 integration).
Co-Authored-By: nebulacoder-v8.0 <noreply@zte.com.cn>
* test(qa-lab): cover GID and cleanup-timeout CLI reject-before-gateway
* fix(qa-lab): drop stale live-scenario mocks from SUT uid tests
* style(qa-lab): oxfmt import order in telegram cli.runtime
* test(qa-lab): streamline strict SUT integer coverage
---------
Co-authored-by: nebulacoder-v8.0 <noreply@zte.com.cn>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Preserve existing lower-level high effort while making xhigh/max available before runtime activation.
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(anthropic): accept "cli" entrypoint in Claude session catalog discovery
Claude Code v2.x writes entrypoint: "cli" in its JSONL session files.
Previously only "sdk-cli" was recognized, causing all v2.x sessions to
be silently skipped.
Extract CLI_ENTRYPOINTS set and isCliEntrypoint() helper, then apply
the same fix to both TypeScript catalog and macOS paired-node native
catalog. Add matching regression coverage on both platforms.
Closes#105164
* fix(anthropic): refresh Claude CLI session discovery
Apply the reviewed cli/sdk-cli allowlist, mirrored negative coverage, and fallback-scope documentation to current main.
Co-authored-by: 黄攀0668000858 <huang.pan@xydigit.com>
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Josh Avant