Agent Claude Code
Implements or refactors C/C++ against the repo's conventions, verifying with its build/tests/clang-tidy/format gate; returns a change report with verbatim gate output. - Use when writing, refactoring, or fixing C/C++, or when a CMake, Makefile, or meson build or its sanitizers gate a change. Spawn one per module-sized…
Agent Claude Code
Digests a file set, docs or source, into a structured brief against the caller's question template, keeping raw bytes out of the caller's context; sweeps or compares a source tree; coverage mode returns only the gaps for a doc slated for deletion or merge. - Use when only the findings come back, not the files.…
Agent Claude Code
Reconciles README, docs/, CLAUDE.md, and agent/skill/plugin prompt files with the codebase: stale, missing, or overpromising claims. - Use when docs lag the code or a change alters documented behavior. Returns the reconciliation list plus items flagged for a decision. Edits docs only. Spawn one per repo.
Agent Claude Code
Implements or refactors Flutter/Dart against the repo's conventions, verifying with its codegen/format/analyze/test gate; returns a change report with verbatim gate output. - Use when writing or refactoring Flutter/Dart: widgets, screens, state management, or buildrunner codegen. Spawn one per module-sized task. Not…
Agent Claude Code
Reviews Flutter/Dart diffs read-only on source and returns a findings table: correctness, async/context safety, rebuild perf, state-management misuse. Writes only its report, to a caller-named path outside the repo. - Use after a Flutter/Dart change lands, before it merges. Spawn one per diff. Not for implementing…
Agent Claude Code
Builds self-contained single-file frontends: canvas/WebAudio demos, inline-SVG icons/favicons, no-build client-side pages. - Use for a standalone artifact: a demo, an icon, a favicon, or a no-build page. Writes frontend files only; returns the files written plus pinned vendored versions. Spawn one per artifact.
Agent Claude Code
Implements or refactors Go against the repo's conventions, verifying with its build/test/vet or golangci-lint gate; returns a change report with verbatim gate output. - Use when writing, refactoring, or fixing Go: packages, goroutines, channels, or context. Spawn one per module-sized task.
Agent Claude Code
Tests a Flutter or mobile app like a real user on an Android AVD or iOS simulator, verifying with parsed screenshots and integration tests. - Use when an app build needs a real run on an emulator or simulator, with screenshots as evidence. Read-only on source; reports pass/fail with evidence paths. Spawn one per app…
Agent Claude Code
Runs a build, test, lint, or any command and returns a pass/fail probe (exit code, warn/error counts) instead of the full log. - Use when only the pass/fail outcome matters, not the log, or to double-check a gate result. Never mutates; no analysis, no fixes. Spawn one per check.
Agent Claude Code
Implements or refactors Python against the repo's conventions, verifying with its test/lint/type-check gate; returns a change report with verbatim gate output. - Use when writing, refactoring, or fixing Python, or when pytest, ruff, or mypy gate a change. Spawn one per module-sized task. Not for reviewing a diff…
Agent Claude Code
Reviews Python diffs read-only on source and returns a findings table: correctness, typing, security, clarity. Writes only its report, to a caller-named path outside the repo. - Use after a Python change lands, before it merges. Spawn one per diff. Not for implementing (python-pro).
Agent Claude Code
Implements, refactors, or tests Rust TUI code with current-ratatui built-ins, writing TestBackend tests and running the repo's gate. - Use for ratatui or crossterm widgets, layout, event loops, or a render bug. Edits Rust source plus the target repo's ratatui punch-list; returns a change report with gate output. Spawn…
Agent Claude Code
Evidence-first bug diagnosis; builds repro harnesses, runs A/B and version-matrix experiments, returns a hypothesis ledger with an attributed cause. - Use when a failure has multiple suspects, a 'started failing recently' claim, or symptoms across unrelated components. Never fixes or posts; writes only its harness.…
Agent Claude Code
Implements or refactors Rust against the repo's conventions, verifying with its build/test/clippy/fmt gate; returns a change report with verbatim gate output. - Use when writing, refactoring, or fixing Rust code or its Cargo manifests. Spawn one per module-sized task. Not for reviewing a diff (rust-reviewer).
Agent Claude Code
Reviews Rust diffs read-only on source and returns a findings table: correctness, safety, async, invariants. Writes only its report, to a caller-named path outside the repo. - Use after a Rust change lands, before it merges. Spawn one per diff. Not for implementing (rust-pro).
Agent Claude Code
Writes or refactors shell scripts (bash or POSIX sh) matching the repo's portability posture, verifying with shellcheck and the repo's tests. - Use when writing or hardening a bash or POSIX sh script, or when shellcheck flags one. Spawn one per script or module-sized task. Not for reviewing a diff (shell-reviewer).
Agent Claude Code
Reviews shell scripts read-only on source and returns a findings table: quoting, error handling, injection, portability. Writes only its report, to a caller-named path outside the repo. - Use after a shell change lands, before it merges; covers scripts embedded in CI yaml or Dockerfiles. Spawn one per diff or script…
Agent Claude Code
Folds a decided spec or approved decision into one design doc that doesn't mention it yet, matching the doc's voice. - Use when a new decision or ADR lands that an existing design doc doesn't carry yet. Returns a short reconciliation report. Edits docs only. Spawn one per doc.
Agent Claude Code
Architecture-level threat modeler (STRIDE, attack trees, requirement extraction): identifies threats, designs mitigations for a system or feature. - Use when a system or feature needs a security review of its design. Read-only on code; writes only the threat-model doc; returns its path plus a prioritized threat…
Agent Claude Code
Reviews TypeScript/JavaScript diffs read-only on source and returns a findings table: correctness, type-safety, XSS/injection, any leakage. Writes only its report, to a caller-named path outside the repo. - Use after a TypeScript or JavaScript change lands, before it merges. Spawn one per diff.
Agent Claude Code
Drives a TUI or CLI program like a real user in a local tmux pty. - Use when a TUI or CLI needs a real run in a terminal: black-box, drives the built binary, never its unit tests. Read-only on source; builds or launches the target itself, reports pass/fail with captured screens. Spawn one per program or flow.
Agent Claude Code
Implements a screen's interaction behavior with the repo's own components: loading/success/error/empty/partial states, form validation, error placement, accessibility wiring. - Use when a screen misses non-happy-path states or a flow needs error handling. Returns a five-state inventory with verbatim gate output. Spawn…
Agent Claude Code
Reviews UI behavior read-only on source and returns a findings table: missing loading/error/empty/partial states, silent failures, error wording, validation gaps, no undo, status-message accessibility. Writes only its report, to a caller-named path outside the repo. - Use after UI behavior changes land, before they…
Agent Claude Code
Single-topic web researcher: multi-query search, full-source fetch, cross-verification, a cited markdown brief returned inline. - Use when one research question needs the web and only the brief comes back; no interaction, no follow-up questions. Spawn one per topic.
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: