Hunt for previously unknown correctness defects in an existing repository by deriving invariants, prioritizing high-risk surfaces, generating concrete bug candidates, and proving or retiring them with bounded evidence. Use when the user asks to find bugs, hidden defects, races, lifecycle failures, data-loss paths, or…
Control dependency upgrades, framework migrations, schema or API transitions, compatibility windows, staged rollouts, rollback, and removal of legacy paths using current evidence. Use when a repository must move from a known current state to a defined target state without breaking consumers. Do not use to diagnose an…
Discover and prioritize evidence-backed opportunities to improve an existing codebase when no specific change has been chosen yet, then recommend the highest-value next upgrade and optionally execute an approved bounded maintenance slice. Use when the user asks what should be improved, modernized, simplified…
Inspect and sanitize provenance or metadata surfaces in user-owned text, images, PDFs, and document containers through an optional local watermarks-remover service. Use when the user wants evidence-first removal of invisible Unicode, C2PA, EXIF/XMP, or document metadata. Do not use for detector evasion, authorship…
Diagnose a concrete failure through controlled reproduction, ranked hypotheses, minimally invasive instrumentation, causal tracing, and falsification, then recommend the smallest explanatory fix and regression test. Use when observed behavior is wrong and the cause is uncertain. Do not use for planned upgrades, broad…
Delegate independent, bounded repository work to specialized reader and writer subagents using mission cards, exclusive write ownership, staged fan-out, and parent-side verification. Use for parallel exploration, reviews, research, isolated implementation, tests, migrations, or security-sensitive work. Do not use when…
Detect documentation drift caused by code, API, schema, configuration, migration, UI, deployment, or operational changes and update all affected user and maintainer surfaces consistently. Use when behavior or contracts changed and documentation must be mapped, edited, and validated. Do not use to invent product…
Build or audit a production Expo and React Native application across Expo Router, development builds, native modules, state/data, persistence, offline behavior, animation/gestures, accessibility, testing, performance, security, observability, EAS builds/updates, and store release. Use when Expo/React Native is…
Build or audit a production Flutter application across architecture, state, routing, networking, serialization, persistence, offline-first behavior, platform integration, adaptive UI, accessibility, testing, performance, security, observability, and release. Use when Flutter is selected or already established. Do not…
Select and shape an evidence-based mobile approach among Flutter, Expo/React Native, native iOS/Android, Kotlin Multiplatform, web-container, and other viable options using product, team, device, security, offline, performance, accessibility, and distribution constraints. Use when platform choice or mobile…
Plan and govern safe parallel repository work using a dependency graph, exclusive write ownership, overlap detection, integration order, and evidence-based acceptance of subagent results. Use when multiple bounded work items can proceed independently after repository boundaries are known. Do not use for repository…
Measure, diagnose, propose, and execute bounded performance improvements for latency, throughput, CPU, memory, GPU, I/O, startup, database, bundle, rendering, or infrastructure cost while preserving correctness. Use when a user asks to profile a critical path, explain a bottleneck, reduce resource use, or optimize…
Define or elevate product UX and visual direction through brief inference, brand evidence, user-flow and state reasoning, explicit design calibration, responsive systems, accessibility, and redesign critique. Use when the core decision is what the experience should communicate and how it should feel or behave. Do not…
Build or audit production web interfaces across architecture, rendering, components, data, forms, authentication boundaries, motion, accessibility, performance, responsive behavior, testing, observability, SEO, deployment, and final browser verification. Use when implementation quality and production behavior are the…
Map a repository's architecture, ownership, dependency boundaries, risk hotspots, change impact, and likely edit conflicts with traceable evidence. Use when a task asks what exists, who or what owns it, what a proposed change can affect, or where parallel edits may collide. Do not use to delegate work, implement the…
Review a defined diff or code area for actionable defects and integration risks, assess structural problems, and plan or execute behavior-preserving refactors through traceable evidence and incremental verification. Use when a user asks for code review, change-risk analysis, maintainability assessment, cleanup…
Reconstruct a maintainable interface from screenshots or visual references through evidence capture, structural decomposition, asset handling, responsive inference, component boundaries, accessibility, and iterative visual comparison. Use when fidelity to supplied visual evidence is the primary requirement. Do not use…
Resolve local shell, runtime, package-manager, browser, native-command quoting, and text-encoding constraints before repeated builds, tests, scripts, or browser checks. Use when required executables or invocation behavior are unknown, or when a task reports missing commands, blocked shims, incompatible runtimes…
Audit and strengthen TypeScript/JavaScript type-evidence and lint discipline, stage adoption of deterministic anti-slop Oxlint rules, and remediate approved violations without laundering diagnostics. Use when a repository needs stronger TypeScript/JavaScript quality enforcement, unsafe type escape-hatch cleanup…
Evidence-backed completion discipline for substantial tasks. Use when the user explicitly asks for exhaustive follow-through, every requested item finished, a long autonomous task to continue until measurable completion, or invokes unlazy, depth-tree, or gate-based execution. Do not use for trivial edits, factual…
Build a risk-based verification strategy and decide whether an integrated change is ready to release from traceable test, CI, coverage, operational, security, compatibility, and rollback evidence. Use when the question is what must be proven or whether a completed change can ship. Do not use to implement the feature…