Reusable Claude Code slash command agents for managing software projects. Each agent is a persona with a defined role, decision authority, and context loading strategy.

CEO (user)
  ├── COO          → task sequencing, process, project board
  ├── Muse         → advisory only, no veto
  ├── Architect    → technical decisions (blocking authority)
  │     └── Deployer → infrastructure within approved architecture
  │     └── Marketing → positioning within approved direction (via COO)
  └── Reviewer     → code quality gate (blocking authority before commits/PRs)

cp agents/*.md ~/.claude/commands/

git clone https://github.com/saldistefano/claude-agents ~/code/claude-agents
ln -s ~/code/claude-agents/agents ~/.claude/commands

Copy the agents/ directory into your dotfiles structure and symlink as appropriate.

Once installed, invoke any agent with its slash command in Claude Code:

/coo                      → get current project status, blockers, next actions
/muse                     → challenge current direction, generate ideas
/architect                → validate a technical decision before implementing
/deployer                 → plan a deployment, pipeline, or infrastructure change
/marketing                → draft a post, plan a launch, review positioning
/reviewer                 → review staged code before committing (standard mode)
/reviewer --ride-along    → walk through code together; teaches Rust patterns as it reviews

These agents are designed to load project context from a kb/ directory following the Meridian knowledge base convention.

Each agent reads specific kb/*.md files on startup. If your project uses a different structure, edit the "Context I Load on Startup" section of each agent file.

• CEO is always the human. Agents advise, inform, and implement. They never override the user.
• Architect has blocking authority — nothing gets built until it's technically approved.
• Blog diary at end of every session — marketing and coo both remind you to write it.
• Agents are project-agnostic — edit the context-loading section for your project structure.

references/coding-guide.md is a mirror of the canonical file at dotfiles/claude/references/coding-guide.md. When adding new principles, edit dotfiles first, then sync here.

The /reviewer agent checklist is backed by primary research. See references/reviewer-references.md for the full annotated source list covering:

• Official Rust Style Guide and API Guidelines
• Google eng-practices code review standard
• Conventional Comments specification
• Rust async pitfalls (Alice Ryhl, Qovery, Google Comprehensive Rust)
• OWASP SSRF and security review guides
• Error handling patterns (Luca Palmieri)
• Privacy and observability (secrecy crate, tracing)

Add new references there as the checklist evolves.

These agents were built for the Meridian project. PRs welcome for improvements, new agent types, or project-specific variants.