P/07
Building with Claude
8 in the head · 7 extras · library of Oct 20, 2026
Suggested when the project contains a .claude-plugin/, a hooks/hooks.json, skills/*/SKILL.md files or a marketplace.json, up to 3 levels deep
Pack head
in order, to load together
Possible extras
depending on the project, on top of the head
- 01
optimize-loop
gaasher/Agent-Loop-Skills
measures each change, keeps it if tests pass , which improves: simpler code or faster SQL
score 8/10high risk175 repo stars - 02
agent-security-hardening
affaan-m/ECC
validates credentials, paths and URLs before each agent action , which improves: blocks leaks and out-of-scope writes
score 9/10no risk284.5K repo stars - 03
build-mcpb
anthropics/claude-plugins-official
bundles a local MCP server with its runtime , which improves: installable without Node or Python
score 10/10no risk43.3K repo starstrusted publisher - 04
deepeval
confident-ai/deepeval
guides pytest evaluation of AI agents , which improves: spots failures and improves their reliability
score 9/10high risk19K repo stars - 05
cookbook-audit
anthropics/claude-cookbooks
scores a Cookbook notebook with its grid , which improves: improves clarity and technical accuracy
score 6/10high risk53.6K repo starstrusted publisher - 06
expo-skill-feedback
expo/skills
sends targeted feedback on an Expo skill , which improves: helps Expo fix what blocks users
score 7/10high risk3K repo starstrusted publisher - 07
aas-compose-stack
sickn33/agentic-awesome-skills
composes a validated AAS skill stack , which improves: keeps a verifiable trace without automatic installation
score 5/10no risk48K repo stars