Frad’s Claude Code Plugins
Install focused Claude Code workflows for engineering, research, content production, and productivity.
- Source repo
- FradSer/dotclaude
- Stars
- ★ 593
- Last updated
- 3d ago
- License
- MIT
- Primary language
- JavaScript
- FA score
- 45/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificClaude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer using Claude Code who wants conventional, atomic commits backed by repository health checks.
- Not a fit if
- Teams that do not use Claude Code
- Users seeking a standalone app or self-hosted service
- Users who need every feature to work without external keys or tools
- Source review
- 45/100 · Major gaps
What does this agent do, and when should you use it?
dotclaude is a collection of Claude Code plugins spanning Git and GitHub automation, refactoring, BDD, SwiftUI review, security checks, research, content production, and hardware design. Its deployable units are separate plugins installed with `claude plugin install <name>@frad-dotclaude`, not a single application or background service. Plugins expose agents, skills, and slash commands, with documented entry points such as `/git-agent:commit`, `/github:review-pr`, `/memory:consolidate`, and `/pi:delegate`. Outputs vary by plugin and include atomic commits, PR review findings, research articles, patent applications, videos, images, PCB fabrication files, and STL, DXF, or PNG assets. The primary runtime boundary is Claude Code, while selected plugins call external CLIs, search services, model endpoints, or remote agents and may require their own credentials.
A user selects a plugin, installs it, and invokes its skills, agents, or slash commands from Claude Code. git handles conventional commits, repository management, and GitFlow branches; git-agent uses @git-agent to inspect gitignore completeness, commit-scope freshness, TruffleHog secret leaks, and co-change index health. github adds quality gates, TDD workflows, validation, and persistent CI and reviewer-comment monitoring through /github:review-pr, while refactor and swiftui address cross-file refactoring and SwiftUI Clean Architecture review. Research and content plugins run autoresearch iteration loops, STORM-style retrieval and article generation, image generation, Remotion or HyperFrames video work, and marketing or office-document workflows. Other plugins operate through Gemini Managed Agents, pi, the Feishu/Lark CLI, KiCad 9.0, OpenSCAD, DeepWiki, Context7, Exa, Git clone, and web search to produce task-specific results and artifacts.
- A developer using Claude Code who wants conventional, atomic commits backed by repository health checks.
- A GitHub project team that needs ongoing PR CI monitoring and triage of reviewer comments.
- An engineering team practicing BDD or TDD that wants checklists, validation, and workflow orchestration inside Claude Code.
- An iOS or macOS developer reviewing SwiftUI code against Clean Architecture and documented best practices.
- A researcher or technical writer running a resumable, multi-perspective retrieval, outlining, and long-form writing process.
- A hardware designer using KiCad 9.0 or OpenSCAD CLI to produce fabrication files, validation reports, or 3D models.
How do you install or deploy this agent?
Claude Code v0.6+ is required. The source provides direct installation commands for individual plugins but does not document the command for initially adding the frad-dotclaude marketplace; that exact prerequisite is therefore missing for an unconfigured environment. Install the plugin you need, for example:
claude plugin install git@frad-dotclaudeInstall the GitHub workflow plugin:
claude plugin install github@frad-dotclaudeInstall the repository scanning and atomic-commit plugin:
claude plugin install git-agent@frad-dotclaudeIndividual plugins can impose additional requirements. antigravity, for example, explicitly requires GEMINI_API_KEY and uv, while the hardware workflows invoke KiCad 9.0 or the OpenSCAD CLI.
How do you use this agent?
After installation, invoke the selected plugin’s agent, skill, or slash command in a Claude Code session opened for the target workspace. These are documented starting points, although the source does not provide argument examples for them:
@git-agentThis runs the full Git health scan. Atomic commits, related-file queries, and workspace initialization use:
/git-agent:commit
/git-agent:related
/git-agent:initMonitor GitHub PR checks and reviewer comments with:
/github:review-prConsolidate memory manually or delegate a task to pi with:
/memory:consolidate
/pi:delegateBefore using a plugin backed by an external service, satisfy its documented prerequisites; for example, antigravity requires GEMINI_API_KEY and uv.
What are this agent's strengths and limitations?
- Plugins are separated by responsibility, so users can install only the Git, GitHub, refactoring, research, hardware, or interface workflows they need.
- Engineering coverage is concrete: GitFlow, semantic versioning, TDD and BDD, persistent PR monitoring, and TruffleHog secret-leak checks are all represented.
- The collection extends beyond code into KiCad and OpenSCAD outputs, image and video generation, long-form research, and active memory workflows.
- Several plugins expose named slash commands that can be triggered directly from a Claude Code session.
- The core delivery model is tied to Claude Code, with no evidence that the plugins run unchanged in ChatGPT, Codex, or a standalone service.
- The source omits the command for adding the
frad-dotclaudemarketplace, leaving a gap in clean-environment onboarding. - Dependencies vary substantially: Antigravity needs
GEMINI_API_KEYanduv, while hardware workflows need KiCad 9.0 or OpenSCAD. - Plugins can involve external models, search, CLIs, and remote sandboxes, but no unified cost, permission, data-handling, or failure policy is documented.
- Repository metadata is inconsistent: the badge reports 23 plugins while the introductory sentence says 22.
How does this agent compare with similar options?
Some plugins derive from or bridge to named projects. mattpocock forks mattpocock/skills v1.2.3 and adds a self-improving checklist subsystem; storm is a Claude-native port of Stanford STORM’s two-stage research-to-article pipeline; and pi delegates execution or read-only review to dev/pi rather than replacing Claude Code. Their distinguishing value here is packaging those workflows as installable Claude Code components instead of using the upstream projects in their original form.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Frad’s Claude Code Plugins This agent | 45 · Major gaps | Agent plugin / skillFree | ★ 593 | 3d ago | JavaScript | Claude Code |
| Workflow Plugin: Claude Workflow V2 | 50 · Major gaps | Agent plugin / skillFree + model costs | ★ 1.4k | 3mo ago | Python | Claude Code |
| OpenCode Swarm | 74 · Some gaps | Agent plugin / skillFree + model costs | ★ 478 | 1d ago | TypeScript | — |
| ECC Engineering Harness | 66 · Some gaps | Agent plugin / skillFreemium | ★ 268k | 2d ago | JavaScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The evidence shows localized safety measures: the git-agent hook restricts raw commits and standalone staging, memory claims privacy scrubbing, git-agent advertises secret scanning, autoresearch asks about decisions, and Antigravity discloses an API key and remote sandbox. Deductions apply because these controls cover only selected plugins; there is no repository-wide least-privilege policy, confirmation model, data-flow inventory, credential lifecycle, destination list, external-write boundary, or recovery procedure. CI dependencies and GitHub Actions use floating versions without visible locking, vulnerability scanning, or supply-chain mitigation. Some forks and mirrors are attributed, but provenance is not consistently documented for every plugin.
CI discovers plugins, validates manifests, and runs pytest where tests exist. The supplied tests cover command-matching edges, operation without jq, and package decoupling. However, plugins without tests are explicitly skipped, the README's “23” badge conflicts with its “22 plugins” statement, and the evidence provides neither a collection-wide dependency inventory and compatibility matrix nor a consistent failure-message standard. Runtime reliability cannot be inferred from this static review.
The README identifies concrete audiences and scenarios across Git, GitHub, refactoring, research, office work, hardware, marketing, and interface design. The git-agent tests demonstrate comparatively precise command triggering and false-match protection. Deductions apply because repository-wide capability boundaries, per-plugin applicability, conflict handling, and fallback behavior are incomplete. Environment requirements are scattered among references to Claude Code v0.6+, uv, API keys, and KiCad 9.0 rather than presented as a unified compatibility guide.
The top-level README is organized by plugin, supplies a consistent installation command for each, and explains the basic structure for adding one. Naming and manifest conventions receive some test coverage. The complete MIT text agrees with the declared metadata, justifying full credit for licensing. Deductions reflect the absence of centralized examples, an FAQ, known limitations, a complete version policy, and a changelog. Maintenance responsibility is inferable only from the copyright name and contribution steps; no maintainer contact, support commitment, or release process is stated.
The collection describes directly usable outputs for commits, reviews, research, documents, images, and hardware workflows. Features such as Git command guards and multiple code-context retrieval methods plausibly add value beyond a basic assistant. Most effectiveness claims, however, remain short README summaries without representative inputs and outputs, quality benchmarks, demonstrated labor savings, or disclosure of model, API, remote-execution, and resource costs.
CI and tests corroborate a subset of README claims, including plugin-structure validation, the git-agent commit guard, and separation between git and git-agent packages. Many other capabilities appear only in directory summaries, with no corresponding implementation files, tests, or examples in the supplied evidence. Promotional descriptions also do not systematically distinguish verified behavior from intended design or inference, limiting traceability and fact-inference separation.
- This is a low-confidence static review; no plugins, tests, or remote services were executed, so runtime results, determinism, and actual correctness are unconfirmed.
- Several plugins can commit code, operate on GitHub, invoke remote agents, generate media, or access collaboration systems, but the supplied material defines no unified confirmation, permission, or external-data-transfer policy.
- The git-agent hook explicitly fails open for empty input, malformed JSON, or a missing command field, which may bypass the intended commit guard.
- CI skips plugins that lack a test directory, while dependency installation and Actions references are not pinned to immutable digests; the presence of CI should not be treated as repository-wide security or test coverage.
- The README inconsistently reports 23 versus 22 plugins. Verify the actual marketplace manifest and each plugin's dependencies, permissions, and provenance before installation.
FAQ
Must I install the whole collection?
Can these plugins run directly in ChatGPT, Codex, or a generic terminal agent?
Do all plugins work without API credentials?
antigravity explicitly requires GEMINI_API_KEY; the image plugin also connects to Gemini or an OpenAI-compatible endpoint, although credential setup is not documented in the supplied material.