aicraft

Use when auditing or profiling complex Rust CLIs, libraries, Tokio services, Tauri backends, Cargo workspaces, local-first single binaries, or SQLite-backed applications, especially for toolchain baselines, crate boundaries, structural lifecycle, ownership, errors, concurrency, CPU and memory evidence, database migrations and query plans, unsafe or FFI safety, quality gates, and architecture-document alignment.

Use when the user asks to prepare a local ChatGPT review package or explicitly send, capture, or iterate an external ChatGPT review of repository work across standard chats, Projects, and Codex, including multi-round review, capability-aware browser routing, and verified local fixes.

Use when reviewed local repository changes need Git delivery that stops before pull-request creation: path-limited staging, verification, local commit, current-branch push or sync, squash-to-main, temporary branch cleanup, or proof that local and remote refs match.

Use when future codebase work needs a plan before implementation: split requirements into scoped executable tasks with owners, dependencies, validation gates, reject criteria, and auditable subagent coordination.

Use when diagnosing bugs, failing tests, build failures, unexpected behavior, regressions, flaky behavior, or performance problems to reproduce the exact symptom, isolate variables, test falsifiable hypotheses, and confirm the root cause before implementation.

Use when drafting, rewriting, diagnosing, or adapting source-grounded Chinese personal and technical writing, factual project or product copy, and developer-community posts while preserving evidence, voice, disclosures, and technical meaning.

Use when implementing or modifying frontend UI, routes, components, forms, tables, dashboards, responsive behavior, DOM/CSS structure, layout ownership, scroll boundaries, or frontend architecture while preserving the existing stack, design system, state, API contracts, and verification flow.

Use when implementing, porting, modifying, or refactoring Rust APIs, crates, services, CLIs, libraries, Tauri backends, Cargo workspaces, FFI, unsafe code, ownership, errors, async behavior, tests, rustdoc, formatting, or Clippy alignment while preserving repository contracts.

Use when directly operating or verifying browser pages, or collecting evidence for an already-isolated browser-layer failure. Do not use for unexplained or cross-system root-cause diagnosis.

Use when directly operating or verifying a specified real desktop client, or collecting evidence for an already-isolated client-layer failure. Do not use for unexplained or cross-system root-cause diagnosis.

Use when the user explicitly asks to ground repository work, map real files, commands, entry points, project docs or classes, preview AGENTS/project-map content, inventory existing page/component or Rust interface reuse candidates, or compare docs and code before implementation.

Use when performing a read-only independent review of an entire repository, branch comparison, commit range, pull request, release candidate, or prepared review package, especially when findings must be coordinated across frontend, Rust, security, tests, CI, documentation, and repository structure with P0-P3 severity and exact remediation evidence.

Use when auditing complex or cross-cutting frontend architecture across React, Vue, TypeScript, Vite, router/state/component systems, high-density consoles, and Tauri desktops, or when code-review or repo-review delegates a bounded frontend specialist review, with explicit profile selection for architecture, state/data, layout, accessibility, performance, or desktop boundaries.

Use when a known code change or scoped API, config, dependency, upload/download, logging, auth, authorization, token/session, CORS/CSRF, secret, privacy, or release surface needs a lightweight evidence-grounded security audit rather than a repository-wide or deep vulnerability scan, including when code-review or repo-review delegates an explicit security surface for a read-only specialist subreview.

Use when existing local git changes need pre-commit review: classify dirty-tree ownership, inspect diffs and contract chains, select review depth, coordinate scoped specialist subreviews, and propose safe commit groups, exact staging plans, or commit messages without mutating Git state.