Softaworks Agent Skills
A curated collection of skills, agents, and commands to supercharge AI coding agents like Claude Code across development, documentation, planning, and professional workflows.
- Source repo
- softaworks/agent-toolkit
- Stars
- ★ 2.6k
- Last updated
- 7mo ago
- License
- MIT
- Primary language
- Python
- FA score
- 51/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificClaude CodeChatGPT · Codex · Claude.ai (Unsupported)
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A Claude Code user needs to quickly create a Mermaid flowchart for a new feature design. They install the mermaid-diagrams skill and ask the agent to generate it.
- Not a fit if
- Teams not using Claude Code, which cannot use the sub-agents or slash commands
- Teams expecting uniform skill quality, as skills vary and must be evaluated individually
- Source review
- 51/100 · Major gaps
What does this agent do, and when should you use it?
Softaworks agent-toolkit is an open-source repository of 'agent skills' following the Agent Skills format, designed to extend the capabilities of AI coding agents such as Claude Code. It includes over 40 skills covering AI tool integrations (Codex, Gemini, Perplexity), meta-skills (e.g., skill-judge), documentation generation (C4 architecture, Mermaid diagrams), frontend development (React, Material-UI), planning (gepetto), professional communication, testing (QA test planners), and utilities (Jira, Datadog). Skills are packaged as installable plugins via npx or Claude Code's plugin marketplace, and each skill can be installed individually. The repository also offers six specialized sub-agents (e.g., UI/UX designer) and seven slash commands (e.g., /codex-plan), exclusive to Claude Code users. MIT-licensed with contribution guidelines.
The repository provides packaged skills, agents, and commands for AI coding agents. Each skill is a directory with a SKILL.md (including YAML frontmatter) and optional scripts, enabling agents to perform specific tasks like generating Mermaid diagrams, refactoring instruction files, creating design systems, or planning QA tests. After installation, agents can invoke these skills—for instance, the codex skill calls GPT-5.2 for advanced code analysis, or the jira skill allows natural language interaction with Jira. For Claude Code, users can register the repository as a plugin marketplace and install sub-agents (e.g., ascii-ui-mockup-generator) or slash commands (e.g., /codex-plan for implementation planning). Manual installation copies skill directories to ~/.claude/skills/. Outputs are agent-executable instructions and scripts, resulting in documents, diagrams, plans, and other artifacts.
- A Claude Code user needs to quickly create a Mermaid flowchart for a new feature design. They install the mermaid-diagrams skill and ask the agent to generate it.
- A development team wants to remove AI-sounding phrases from generated documentation. They use the humanizer skill to make text more natural.
- A project manager uses the jira skill to create and update Jira tickets via natural language commands, avoiding manual UI interaction.
- A frontend developer needs to convert an OpenAPI specification into TypeScript types for type-safe API calls, using the openapi-to-typescript skill.
- A product manager uses planning skills like gepetto to create detailed implementation plans ensuring requirements are clear before coding.
- A developer finishes a long session and uses the session-handoff skill to seamlessly transfer context to a new AI session.
How do you install or deploy this agent?
- Quick install (recommended for multiple agents): run
npx skills add softaworks/agent-toolkit. - For Claude Code, register as plugin marketplace:
- Run /plugin marketplace add softaworks/agent-toolkit then /plugin.
- Use the browse UI to select plugins, or install directly via commands like /plugin install codex@agent-toolkit.
- Manual installation (skills only):
cp -r skills/<skill-name> ~/.claude/skills/.
Note: Agents and commands require Claude Code CLI.
How do you use this agent?
After installation, you use skills within Claude Code:
- Skills: ask the agent to use a specific skill, e.g., 'Generate a flowchart using the mermaid-diagrams skill.'- Agents: after installing an agent (e.g., /plugin install agent-codebase-pattern-finder@agent-toolkit), Claude Code can delegate tasks to it.
- Commands: after install, invoke slash commands like /codex-plan to create an implementation plan.
- Updates: run /plugin, go to Marketplaces tab, select agent-toolkit, and choose 'Update marketplace' or enable auto-update.
What are this agent's strengths and limitations?
- Extensive skill library covering development, documentation, planning, testing, and more, directly boosting daily efficiency.
- Follows the Agent Skills standard, making it easy to extend and contribute.
- Supports granular installation—install only the skills, agents, or commands you need, avoiding bloat.
- Plugin marketplace integration simplifies management and updates.
- Agents and commands are exclusive to Claude Code; other platforms cannot use them.
- Some skills depend on external APIs (e.g., GPT-5.2, Gemini, Perplexity, Jira, Datadog), potentially incurring costs or requiring API key setup.
- Skill quality varies—users must evaluate each skill (the skill-judge skill helps with this).
- Manual installation only supports skills, not agents or commands, limiting functionality.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Softaworks Agent Skills This agent | 51 · Major gaps | Agent plugin / skillFree + model costs | ★ 2.6k | 7mo ago | Python | Claude Code |
| Keep the Why | 75 · Good | Agent plugin / skillFree | ★ 231 | 1d ago | Python | Codex · Claude Code |
| cc-skills-golang: Agent Skills for Production-Ready Go | 35 · Major gaps | Agent plugin / skillFree + model costs | ★ 3.5k | 17d ago | Go | ChatGPT · Codex · Claude Code |
| JS Mastery Engineering Workflow Skills | 58 · Major gaps | Agent plugin / skillFree + model costs | ★ 1.5k | 2mo ago | JavaScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows CI workflows with minimal permissions (contents: write for auto-bump), but no explicit user confirmation or data flow transparency. Skills may execute external commands (e.g., datadog-cli, jira) without clear permission boundaries or user prompts. Dependency security not mentioned, external effects unclear, rollback not documented, source attribution only via README author link. Thus most criteria scored 1 due to insufficient evidence or partial presence.
Repository structure is consistent, skill format uniform, but no failure messages or error handling documented. Dependency availability not verified, only relies on npx and Python without version pinning. Self-consistency good as README and skill structure align.
Clearly targets AI coding agents like Claude Code, covering development, documentation, planning scenarios. Capability boundaries described via skill descriptions and README table, but trigger conditions not detailed. Environment fit documented with installation methods, but compatibility limitations not mentioned.
Information architecture clear with README navigation and skill table. Installation notes detailed, naming stable (consistent skill names). Examples and FAQ partially present, but known limitations not explicit. License MIT, versioning via auto-bump workflow, maintenance responsibility via contributing guide and author info.
Output usability evidenced by skill descriptions and examples, marginal value in providing diverse skills, cost-benefit good as free open-source, but no performance or resource consumption data.
Claim traceability moderate, README descriptions lack detailed evidence. Cross-source corroboration insufficient, facts and inferences not clearly separated.
- Skills may execute external commands (e.g., datadog-cli, jira) without clear permission boundaries or user confirmation, posing potential risks.
- Dependency security not addressed; dependency versions or vulnerabilities not verified.
- No rollback mechanism or failure recovery documented.