Keep work dependable
Adapt an interface without losing work
Choose ordinary DOM, framework state, or declarative rendering while preserving drafts, focus, and stable task identity.
- Ordinary DOM, framework state, or a narrow declarative model
- Logical identity versus task/question identity
- In-place reconciliation or a deliberate disruptive update
Choose capabilities for your task · Evidence: explained
adaptation · accessibility · drafts
Adaptation request
Collect evidence
Annotate a particular image or frame
Keep coordinates, identity, and transforms coherent when evidence is resized or replaced.
- Normalized coordinates or source pixels
- Paused frame, explicit tracking, or approximate live cue
- Local annotation drafts and shared committed edits
Choose capabilities for your task · Evidence: explained
images · coordinates · identity
Adaptation request
Collect evidence
Ask a question about reviewed evidence
Bind questions and replies to specific sources and distinguish observation from interpretation.
- Question-led source selection and context
- Observed fact, inference, and unresolved uncertainty
- Stable evidence/reply correlation
Choose capabilities for your task · Evidence: explained
evidence · reasoning · correlation
Adaptation request
Collect evidence
Capture rendered evidence
Choose between DOM rendering, browser capture, and underlying source data, and review what the capture leaves out.
- Rendered appearance or source/data export
- DOM reconstruction or browser/screen capture
- Explicit readiness and reviewed disclosure
Choose capabilities for your task · Evidence: explained
capture · rendering · provenance
Adaptation request
Collect evidence
Capture, review, and send evidence
Design permission, review, durable submission, and recovery around evidence the person actually intends to send.
- File or typed fallback versus a live sensor
- Inline review versus a separate capture route
- Local draft lifetime and late rejection behavior
Selected capabilities: captures, durable commands · Evidence: exercised
capture · privacy · drafts
Adaptation request
Keep work dependable
Choose the right home for application data
Match authority, ordering, lifetime, and privacy to each kind of information instead of putting everything in shared state.
- Current value versus ordered history
- Authority, privacy, and record lifetime
- View checkpoints versus handled-agent-work checkpoints
Choose capabilities for your task · Evidence: explained
data-model · authority · retention
Adaptation request
Guide and coordinate work
Choose when to interrupt
Match quiet telemetry, agent wake, local cues, timers, and notifications to a meaningful decision point.
- Quiet record, agent wake, local cue, timer, or push
- Coalescing, cooldown, and acknowledgement meaning
- Missed-time grace, time zones, and deduplication
Choose capabilities for your task · Evidence: explained
notifications · attention · timers
Adaptation request
Choose inputs and algorithms
Choose a voice interaction
Compare typed input, push-to-talk, energy detection, and explicit wake engines with consent and reliable stop behavior.
- Typed input, push-to-talk, VAD, or wake engine
- Local processing and explicit resource installation
- Review recognized text and control output independently
Selected capabilities: optional voice adapter · Evidence: explained
voice · consent · engines
Adaptation request
Guide and coordinate work
Collect private responses and publish a reviewed aggregate
Keep answers and assignments private while giving a facilitator an explicit, versioned path to shared results.
- Public poll or authorized private records
- Revisioned answers or append-only revisions
- Reviewed aggregate and atomic round reset
Selected capabilities: explicit trusted extension, private records, transactions · Evidence: exercised
privacy · collaboration · transactions
Adaptation request
Collect evidence
Collect and discuss a sketch
Choose a drawing model, preserve local work, and submit a reviewed snapshot with meaningful context.
- Bitmap snapshot or editable stroke/object model
- Local drawing versus live shared editing
- Resize, review, and pending-snapshot continuity
Choose capabilities for your task · Evidence: explained
canvas · drafts · review
Adaptation request
Decide and act
Build a comparison and decision workspace
Make criteria, units, provenance, and uncertainty inspectable instead of disguising assumptions behind a single score.
- Table, small multiples, narrative, or interactive weights
- Hard constraints versus compensating preferences
- Missing evidence and sensitivity to assumptions
Choose capabilities for your task · Evidence: explained
comparison · decisions · uncertainty
Adaptation request
Choose inputs and algorithms
Connect an instrument and interpret its protocol
Separate permission, packet framing, command policy, and bounded telemetry from a task-specific hardware interface.
- Direct device, remote collector, or uploaded/simulated data
- Packet framing and bounded carry buffers
- Read-only access versus explicit physical commands
Choose capabilities for your task · Evidence: explained
hardware · protocols · telemetry
Adaptation request
Keep work dependable
Connect an agent that actually resumes
Choose and verify a host receiver that wakes the intended conversation and acknowledges only handled work.
- Socket/IPC, wait task, webhook, or scheduled check
- Exact receiver identity and count-only wake notices
- Delivered versus handled checkpoints and rearming
Choose capabilities for your task · Evidence: explained
agents · delivery · recovery
Adaptation request
Guide and coordinate work
Coordinate views and attention across devices
Assign useful device purposes while keeping authority, freshness, focus, and completion receipts distinct.
- Same view or complementary device purposes
- Targeted display versus authorized private access
- Audio focus, freshness, presentation, and completion
Choose capabilities for your task · Evidence: explained
devices · attention · handoff
Adaptation request
Choose inputs and algorithms
Detect meaningful changes cheaply
Choose baseline, hysteresis, debounce, and cooldown according to the cost of false triggers and missed events.
- Threshold, hysteresis, debounce, or cooldown
- Fixed or adaptive baseline and first observation
- False-trigger versus missed-event costs
Choose capabilities for your task · Evidence: explained
heuristics · hysteresis · sampling
Adaptation request
Choose inputs and algorithms
Discover capabilities and design around them
Distinguish API presence from usable permission and offer a task-completing alternative before requesting access.
- No-prompt observation versus a relevant permission request
- API presence, permission, hardware, and actual operation
- Fallbacks that complete the same task
Choose capabilities for your task · Evidence: explained
capabilities · permissions · fallbacks
Adaptation request
Choose inputs and algorithms
Follow location changes and route progress
Filter uncertain fixes, choose arrival semantics, and keep a route usable when foreground location stops.
- One fix, manual progress, or continuous tracking
- Accuracy filtering and inner/outer arrival radii
- First-fix, skipped-waypoint, and replanning semantics
Choose capabilities for your task · Evidence: explained
location · uncertainty · routes
Adaptation request
Guide and coordinate work
Guide a task with explicit checkpoints
Separate task definition from progress, and choose intentional completion, resume, and replanning behavior.
- Checklist, focused step, or non-linear dependency view
- Manual completion, observed evidence, or reviewed inference
- Resume and replanning across definition revisions
Choose capabilities for your task · Evidence: explained
guidance · progress · replanning
Adaptation request
Collect evidence
Import information for review
Turn pasted or uploaded data into a correctable model with preserved originals and explicit partial acceptance.
- Manual entry or staged import
- All-or-nothing versus explicit partial acceptance
- Ambiguous mappings, preserved originals, and retry identity
Choose capabilities for your task · Evidence: explained
import · review · provenance
Adaptation request
Keep work dependable
Ingest telemetry with a bounded budget
Choose acquisition and spool behavior while making backpressure, retention, and data loss explicit.
- File tail, snapshot command, or remote source
- Scoped producer identity and stable retry records
- Bounded spool, backpressure, and explicit loss
Choose capabilities for your task · Evidence: explained
telemetry · spool · retention
Adaptation request
Choose inputs and algorithms
Turn noisy motion into useful events
Calibrate simple motion heuristics and describe their uncertainty before treating an event as progress.
- Manual marking or heuristic events
- Rising edge/refractory period or low/high hysteresis
- Raw diagnostics, corrections, and summary uncertainty
Choose capabilities for your task · Evidence: explained
motion · calibration · heuristics
Adaptation request
Guide and coordinate work
Join or transfer a page across devices
Choose role-aware links and a usable transfer method without turning device purpose into authority.
- QR, typed link, or explicit device handoff
- Same page, participant view, or shared projection
- Authentication and draft-transfer policy
Choose capabilities for your task · Evidence: explained
devices · links · roles
Adaptation request
Keep work dependable
Keep an active task usable
Choose foreground wake lock and interruption behavior without promising background execution.
- Ordinary screen behavior or explicit active-task wake lock
- Requested versus actual lock state
- Reacquisition, task stop, and background limits
Choose capabilities for your task · Evidence: explained
foreground · lifecycle · wake-lock
Adaptation request
Choose inputs and algorithms
Notice meaningful visual changes
Choose reference frames, thresholds, and sampling budgets for a task where a changed scene matters.
- Previous, uploaded, or fixed reference frame
- Pixel difference or a selected perception model
- Sampling cadence versus upload cooldown
Choose capabilities for your task · Evidence: explained
camera · heuristics · budgets
Adaptation request
Decide and act
Observe long-running work and enable intervention
Separate freshness, progress, failures, and intervention from the agent conversation that may come and go.
- Completion notice or continuous observation
- Observed process states and freshness
- Polling/events, scoped collectors, and deliberate intervention
Choose capabilities for your task · Evidence: explained
processes · status · intervention
Adaptation request
Make sense of information
Present live measurements honestly
Choose a bounded data model, rendering approach, and recovery policy without hiding gaps or mixing runs.
- Latest reading, trend, or run comparison
- Raw values, EMA, or time-aware smoothing
- Replay or an atomic view-and-cursor checkpoint
Selected capabilities: streams · Evidence: exercised
streams · visualization · recovery
Adaptation request
Keep work dependable
Publish coherent page updates
Keep authored assets, runtime identity, nested paths, and draft-aware updates consistent within one target system.
- Provenance, deliberate dependency, and final publication identity
- Nested relative files and explicit routing assumptions
- Draft-aware publication readiness and update behavior
Choose capabilities for your task · Evidence: explained
publication · integrity · updates
Adaptation request
Guide and coordinate work
Publish the smallest useful page
Start with ordinary authored files and one useful event/reply loop, then add structure only when the task needs it.
- One file, modules, or nested pages
- Event/reply interaction or a dashboard without conversation
- Exact authored assets and a verified receiver
Choose capabilities for your task · Evidence: explained
publishing · plain-files · first-page
Adaptation request
Keep work dependable
Maintain a shared editable collection
Choose stable identities and explicit conflict behavior for additions, edits, ordering, and deletion.
- Append-only, field patches, revision guards, or human merge
- Stable identity separate from ordering
- Deletion, optimistic edits, and stale-draft review
Choose capabilities for your task · Evidence: explained
concurrency · drafts · collections
Adaptation request
Keep work dependable
Recover from queued, repeated, and outdated work
Keep original command meaning across interruption and distinguish retry from a new, deliberately reviewed action.
- Queued, accepted, rejected, and completed are distinct
- Retry original context or deliberately compose new work
- Submitted snapshot separate from current draft
Choose capabilities for your task · Evidence: explained
commands · generations · drafts
Adaptation request
Keep work dependable
Resume across devices and agent sessions
Keep task context, consumer progress, view checkpoints, and device focus separate so a handoff does not skip work.
- Task brief, handled cursor, view checkpoint, or local draft
- Authorized cross-device state versus private unsent work
- Receiver identity and local consent after handoff
Choose capabilities for your task · Evidence: explained
handoff · resume · context
Adaptation request
Decide and act
Review an exact proposal before acting
Bind approval to immutable content and an explicit version, expiry, and execution policy.
- Approval or a simpler authorized progress receipt
- Response-time and execution-time expiry
- Downstream idempotency and uncertain outcomes
Selected capabilities: explicit trusted extension · Evidence: exercised
approval · versions · authority
Adaptation request
Make sense of information
Run a bounded measurement experiment
Choose acquisition, aggregation, intervention, and export policies before treating a dashboard as an experimental record.
- Live acquisition, uploaded data, or declared simulation
- Raw retention versus explicit summary buckets
- Intervention records and honest export scope
Choose capabilities for your task · Evidence: explained
experiments · aggregation · provenance
Adaptation request
Choose inputs and algorithms
Add local perception with an explicit engine
Choose model assets, confidence rules, coordinate conventions, and evaluation for optional pose, object, or text recognition.
- Manual interpretation versus a selected model
- Engine assets, provenance, and processing location
- Confidence, frame identity, and bounded inference
Choose capabilities for your task · Evidence: explained
perception · experimental · models
Adaptation request
Collect evidence
Keep evidence and interpretation separate
Preserve original observations while allowing conclusions, labels, and explanations to evolve.
- Append-only note or structured observation record
- Original evidence versus editable interpretation
- Retention, search coverage, and export scope
Choose capabilities for your task · Evidence: explained
evidence · provenance · notes
Adaptation request
Keep work dependable
Test an adaptive page honestly
Use exact authored files and explicit fixtures, then verify authority and transactions at the real service boundary.
- Numerical, browser, publication, or actual-service evidence
- Explicit fixture capability and missing-capability cases
- Behavioral freedom without pixel or component identity
Choose capabilities for your task · Evidence: explained
testing · fixtures · evidence
Adaptation request
Make sense of information
Turn messy records into useful signals
Project fields, preserve units, and bound flattening without silently changing the meaning of a source record.
- Explicit field projection or bounded flattening
- Missing/null/malformed values and units
- Collision, depth, cardinality, and byte limits
Choose capabilities for your task · Evidence: explained
records · projection · limits
Adaptation request
No matching approach. Try a broader outcome or clear the goal filter.