Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/multiplex-term/multiplex/sync-and-widgetsgit clone --depth 1 https://github.com/multiplex-term/MultiplexWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00996 |
| Opus 5 | $0.00000 | $0.00498 |
| Sonnet 5 | $0.00000 | $0.00199 |
| Haiku 4.5 | $0.00000 | $0.00100 |
Grade A, and why
sync-and-widgets scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cross-device sync, widgets & Shortcuts
Load-bearing decisions split from AGENTS.md.
- Cross-device sync rides iCloud Keychain, nothing else
(E2E-encrypted, no entitlement/CloudKit): secrets AND a JSON host
record per host are synchronizable items (services
app.multiplexterm.multiplex/….multiplex.hosts). Every keychain query must passkSecAttrSynchronizable(Any)— omitting it silently matches only device-local items.HostSync.merge(pure, tested): last writer wins byHost.updatedAt; a locally-persisted mirrored-IDs set distinguishes "new local → publish" from "peer deleted → drop". No change notification exists — deck and terminal roots re-merge on scenePhase.active. - Widgets/Shortcuts add no background execution and must stay that
way: they declare no background mode of their own and must not grow
one (the app's single
fetchmode belongs to the agent-alert refresh — see the keep-alive entries inlifecycle-and-attention.md; widgets never connect at all); intents areopenAppWhenRun; widget timelines are.neverwith app-pushed, hash-gated reloads; probe loops gate network work throughBackgroundActivity(see the keep-alive entry inlifecycle-and-attention.md— still no background mode, and a host that never opted in keeps the old.active-only behaviour). The Shortcuts host picker and widget config readSharedStateStore; Open Agent's dependent pickers re-resolve against the liveHostStore. Configured working-dir paths ride widget state for the widget's own directory picker; setup-script names and bodies never do. Directory stays a String for variables (unset = host default,"~"= Home); Open File likewise keeps its remote path as a String, carries an optional positive line number, and recognizes apath:10-15range in that path, registering the in-memory viewer before its route opens; the setup-script String is validated DEFAULT/NONE/UUID; the Model String is gated bynormalizedLaunchModel.SharedStateTestslocks intent/widget/ExternalActionURLformats in lockstep (the widget target compiles onlyMultiplex/Shared— never import Host/Tmux/Agent types there). XcodeGen quirk: the widget target's deployment floors need explicitIPHONEOS_DEPLOYMENT_TARGET/XROS_DEPLOYMENT_TARGET(target-leveldeploymentTargetis ignored for multi-destination targets). E2E without widgets:xcrun simctl openurl <UDID> "multiplex://open?host=devbox&action=shell"; the App Group file lands undersimctl get_app_container … groups. iOS 26 confirms the FIRST simctl-originated open per install. The failure alert presents from the mode root (ExternalActionHost) — never the deck pane, which the expanded shell clips to zero width. - A widget's featured session is the last one OPENED, not the newest
created (
WidgetHostState.featuredSession, pure + tested): the Host widget's own Session setting first (an explicit Backend is strict; Host Default tries the host's default namespace, then wherever the name lives, because the picker lists both backends' names), then the host'slastAttached, thenmostRecentSession(creation date, kept as the final fallback; the bare shell deep link's router fallback follows the same last-opened → newest order). Creation order was the old rule and is meaningless on herdr —HerdrProbesynthesizes near-epoch dates from list order, so a mixed host featured a tmux session forever.HostStore.recentSessions(device-local UserDefaults beside tile order; never synced) is written from ONE place,TerminalWindowUIKit's active-tab chokepoint (first appearance, tab switch, deck/notification reveal) — never fromTerminalWorkspace's controller getter, which runs for every restored tab in array order. The deck observes it likehostsand republishes; the ref rides the snapshot name-only on single-backend hosts and backend-qualified on mixed ones (the row convention); a stale name falls through, never blanks.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 66 lines · 0 tokens per session scan A a2340da9f6e8
sync-and-widgets is an agent published in the GitHub repository multiplex-term/Multiplex (10 stars, last pushed 5d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 996 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
layer3-issue-detection
Layer 3 systematically scans ALL entry points from Layer 1 and applies issue detection rules. Unlike Layer 2 (which traces specific flows in depth), Layer 3 does a breadth-first scan to categorize issues across the entire codebase.
layer4-semantic-evaluation
Layer 4 evaluates workflows from the user's perspective. While Layers 1-3 analyze code structure, Layer 4 asks: "Does this workflow help the user achieve their goal?".
layer2-methodology
Layer 2 takes entry points from Layer 1 and traces the complete user journey.
mobile-architect
Mobile architecture expert. Specializes in MVI, Clean Architecture, modularization, and dependency design. Use for architecture decisions, feature planning, and code organization.
Native SwiftUI Builder
Builds complex custom SwiftUI views using standard SwiftUI components with TTBaseUIKit design tokens (XView/XSize/XFont). Does NOT use TTBaseSUI wrapper components.
kmp-architect
KMP architecture expert. Designs shared module structures, expect/actual patterns, navigation, and DI for Kotlin Multiplatform projects.