Installs, enables, disables, uninstalls, and launches the Antigravity auditor and debugger: a before-tool-call hook plus a local web UI that logs every tool call in real time and can pause a call for inspection without interrupting the session. Use when the user asks to install, enable, disable, remove, start, or open…
Fetches a GitHub pull request's description, reviews, review threads, comments, and CI diagnostics into a JSON file and summarizes what the PR says about itself. Use when the user wants to understand, summarize, or extract information from a PR and its discussion, or when another task needs the PR's context before…
Reviews code against the canonical API design principles: contract-first interfaces, KISS and YAGNI, intent-revealing names, command-query separation, strict types over strings, visible validation, and lean public entry points, and reports findings with concrete rewrites. Use when the user asks for an API review, asks…
Finds git branches whose work has already landed on the main branch, including squash-merged ones, and deletes them along with their worktrees and, on request, their remote tracking branches. Writes an analysis plan first and deletes only what that plan approved. Use when a repository has stale local or remote…
Writes and reviews code documentation: dartdoc, JSDoc, Python docstrings, and implementation comments, with per-language conventions for Dart, Python, and TypeScript. Use when writing new code or features, documenting public APIs, improving existing docstrings or comments, or checking a change for missing or unhelpful…
Reviews code and technical documents (pull requests, local changes, whole files, specs, design docs, RFCs, ADRs, blueprints) in generate, critique, and synthesize passes. Every finding carries an evidence tier and a concrete consequence, and severity is capped by the evidence. Syncs remote PRs into a temporary…
Commits changes to a git repository: checks that the repository is healthy by running the language's formatter, linter, and tests, reviews the staged diff, and commits with a conventional commit message that explains why the change was made. Use when the user asks to commit, "finish up", "wrap this up", or land the…
Guides the structure of Dart command-line applications: a thin bin/ entrypoint, exit code handling with exitCode rather than exit(), exception handling, executable scripts for Linux and macOS, and the recommended packages (args, io, path). Use when creating a Dart CLI app, refactoring a bin/ entrypoint, writing an…
Prevents, diagnoses, and fixes asynchronous problems in Dart and Flutter: UI freezes and jank, event loop starvation, race conditions, shared-state corruption, isolate deadlocks, and test timeouts, using the Dart MCP analysis, test, and runtime tools. Use when the UI hangs or stutters, when a test times out around…
Applies Dart 3 language features when writing or refactoring Dart and Flutter code: records, pattern matching, switch expressions, class modifiers, extension types, digit separators, wildcard variables, null-aware elements, and dot shorthands. Use when new Dart code is being written, when older Dart is being…
Structures Flutter apps along the Flutter team's recommended architecture: MVVM with view models, repositories and services, unidirectional data flow, an optional domain layer of use cases, dependency injection with provider, and ListenableBuilder for reactive UI. Use when starting a new Flutter app, adding a feature…
Moves CPU-bound work in Flutter, such as parsing large JSON payloads, off the UI thread into a background isolate, picks manual or generated JSON serialization, handles the Web platform where isolates are unavailable, and integrates results with state management without leaking resources. Use when a Flutter app janks…
Routes Flutter testing work: picks unit, widget, or integration tests for the task, points to the matching reference guide (mocking, matchers, plugin testing, common errors), and runs tests through the Dart MCP runtests tool with the CLI as fallback. Use when writing or fixing Flutter or Dart tests, choosing a test…
Answers, explains, analyzes, and reviews code using read-only tools only, with no file edits, no file creation, and no commands that change workspace state. Use when the user runs /nochange, says "nochange", or asks to explain, walk through, or review code without editing it. Don't use for short factual lookups where…
Works through review feedback on the user's own GitHub pull request: fetches the unresolved threads, verifies each comment against the code, drafts a fix or reply per thread, collects the user's decisions through a local dashboard or a triage report, then implements the approved fixes with regression tests, pushes…
Recommends primary and secondary reviewers for a GitHub pull request, branch, or commit range from git history, file and directory ownership, blame on the modified lines, and past review collaboration in the repository. Use when asked who should review a change, to find or suggest reviewers for a PR or branch, or to…
Writes technical proposals, RFCs, system designs, and architecture documents for human stakeholders, choosing a lightweight, standard, or detailed template by project size and including non-goals, diagrams, an API review, and a self-critique pass. Use when the user asks for a proposal, RFC, design doc, or architecture…
Inspects a repository through the sem CLI, which tracks functions, classes, and types instead of lines: semantic diffs between commits, dependency and call graphs, transitive impact of changing an entity, semantic blame, and per-entity history. Use when analyzing structural code changes, measuring the blast radius of…
Takes finished code through the pre-review pipeline: splits the work into stackable pull requests, runs static analysis and a test quality audit with 90 percent or higher coverage, loops subagent API and code reviews until clean, syncs code documentation, README, and CHANGELOG, writes the PR description, and tracks…
Creates, edits, and validates agent skills: scaffolds a skill directory with initskill.py, writes the SKILL.md frontmatter as a trigger contract and the body as procedural instructions, organizes scripts, references, and assets, and lints the result with lintskill.py for frontmatter, structure, and link integrity. Use…
Provides portable command-line patterns for text processing, file search, and JSON handling: ripgrep and fd in place of grep and find, BSD versus GNU differences in sed, awk, and xargs on macOS, jq recipes for structured output, parallel execution, and shell one-liner hygiene. Use when writing or reviewing shell…
Writes a pull request description for the current branch from its diff against the base branch: a squash-commit-ready title, the reason for the change, a structured summary of what changed, and the testing performed, saved to the scratch directory for review. Use when the user asks to write, generate, draft, or create…
Enforces clear, natural human writing free of AI-isms, obscure words, em-dashes, inline-header list overuse, and complex sentences. Use when producing, editing, or reviewing any substantial prose deliverable, even if the user does not mention writing style. Applies to artifacts, design docs, reports, READMEs, PR or CL…
Calculates Cognitive Complexity per the SonarSource standard for Python, TypeScript, JavaScript, Dart, Swift, Kotlin, and C or C++ files, with per-function scores, line-by-line increment breakdowns, and JSON output. Use when asked to measure, evaluate, or compare the complexity of code, files, or directories, to find…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: