Vibe Coding Discoveragents-best-practices
View on GitHubProvider-neutral Agent Skill for Codex, Claude Code, and agentic harness design.
Provider-neutral Agent Skill (SKILL.md + reference docs) for designing, auditing, and refactoring agentic harnesses: loops, tool permissions, context/memory compaction, evals, cost and security. Installs into Codex or Claude Code skills directories.
Use Cases
Generate MVP agent harness blueprintsAudit brittle or expensive existing agentsDesign typed tools and permission modelsAdapt agents to late-bound tool environmentsDesign self-refining recursive harnessesSpeculative/parallel tool execution designContext, memory and compaction strategyAgent evals and observability setupPrompt caching and cost controlAgent communication via public boardsSecurity review of agent runtimesRefactor agentic loops and planning
Built With
- Frameworks
- Claude Code · OpenAI Codex · MCP · vercel-labs/skills
Tags
agent-skill · claude-code · codex · agentic-workflows · agent-architecture · harness-design · tool-permissions · context-management · memory-compaction · evals · observability · orchestration · prompt-engineering · mcp · security · cost-optimization