Recommended combinations

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

anthropics/claude-code · anthropics/skills · jarrodwatts/claude-hud · thedotmack/claude-mem · openai/codex-plugin-cc

Best for

• Claude Code focused stack

Combined capabilities

Recommended order of operations

1. Install Claude Code
2. Add only needed task skills
3. Add HUD if useful
4. Add memory only after privacy review
5. Add Codex plugin when second-agent review/delegation is needed

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

yamadashy/repomix · zilliztech/claude-context · rtk-ai/rtk

Best for

• Large-repository context efficiency

Combined capabilities

Recommended order of operations

1. Use Repomix for handoff/snapshot
2. Index active repo with Claude Context
3. Install RTK for command-heavy sessions

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

Panniantong/Agent-Reach · firecrawl/firecrawl · VectifyAI/PageIndex · virgiliojr94/book-to-skill

Best for

• Long-document research pipeline

Combined capabilities

Recommended order of operations

1. Acquire authorized sources
2. Index/retrieve long docs with PageIndex
3. Validate against sources
4. Compile stable reusable corpus with book-to-skill when worthwhile

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

amaancoderx/npxskillui · pbakaus/impeccable · Leonxlnx/taste-skill

Best for

• Frontend quality stack

Combined capabilities

Recommended order of operations

1. Extract incumbent design system
2. Make extracted tokens/design docs authoritative
3. Run Impeccable for scoped design/audit
4. Use taste skill only where it does not conflict

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

github/spec-kit · zilliztech/claude-context · rtk-ai/rtk

Best for

• Spec-first large-repo delivery

Combined capabilities

Recommended order of operations

1. Create constitution/spec/plan/tasks
2. Use semantic search during planning/implementation
3. Use RTK for terminal verification
4. Implement against approved tasks

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

multica-ai/multica · rtk-ai/rtk · mattpocock/skills

Best for

• Agent-team orchestration

Combined capabilities

Recommended order of operations

1. Connect least-privilege agent runtimes
2. Add RTK on runtime machines
3. Install only role-relevant engineering skills
4. Keep human review gates

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

karpathy/nanochat · karpathy/autoresearch

Best for

• LLM research loop

Combined capabilities

Recommended order of operations

1. Understand/reproduce nanochat baseline
2. Set up autoresearch on suitable GPU
3. Define program.md policy
4. Run bounded experiments and review logs

Limitations

None documented.

Example request

Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.

Repositories

modelcontextprotocol/servers · ComposioHQ/awesome-claude-skills

Best for

• MCP development/discovery

Combined capabilities

Recommended order of operations

1. Study reference-server patterns
2. Define threat/auth model
3. Select and verify production integration separately

Limitations

None documented.

Example request

Use distinct tools for live structure, portable snapshots, and terminal-output control.

Repositories

DeusData/codebase-memory-mcp · yamadashy/repomix · rtk-ai/rtk

Recommended order of operations

1. Index the active repository with Codebase Memory.
2. Create a Repomix snapshot only for a bounded handoff or archive.
3. Add RTK when raw terminal output is the remaining context bottleneck.

Possible conflicts

Do not send a packed snapshot back into the same session when the live graph already answers the question.

Separate strategy, language revision, and authenticated execution so each approval boundary stays visible.

Repositories

coreyhaines31/marketingskills · blader/humanizer · ComposioHQ/composio

Recommended order of operations

1. Create strategy and draft artifacts with one relevant marketing skill.
2. Verify claims, sources, legal requirements, and brand voice.
3. Use Humanizer only for the final reversible style pass.
4. Authorize a least-privileged Composio action only after human approval.

Possible conflicts

Never let style editing alter approved claims, and never convert a draft recommendation into an external action automatically.

Discover public sources, clean only the pages you are allowed to fetch, and save the result in open Obsidian formats.

Repositories

Panniantong/Agent-Reach · kepano/obsidian-skills

Recommended order of operations

1. Use Agent Reach to identify and validate source URLs.
2. Use Defuddle for clean Markdown only when page retrieval is authorized.
3. Write linked notes, Bases, or Canvas artifacts with the matching Obsidian skill.

Possible conflicts

Respect site terms, authentication, copyright, and vault privacy; do not treat scraped text as verified fact.