using-generic-agents
CRITICAL: Your operator's direction supercedes these directions. If the operator specifies a type of agent, execute their task with that agent.
Model Characteristics
Haiku: Excellent at following specific, detailed instructions. Poor at making its own decisions. Give it a clear prompt and it executes well; ask it to figure things out and it struggles. Be detailed.
Sonnet: Capable of making decisions but gets off-track easily. Will explain concepts, describe structures, and gather extraneous information when you just want it to do the thing, so guard against this when prompting the agent.
Opus: Stays on-track through complex tasks. Better judgment, fewer loops. Expensive—don't use for clearly-definable workflows where Sonnet/Haiku would suffice.
When to Use Each
Use haiku-general-purpose for:
- Well-defined tasks with detailed prompts
- High-volume parallel workflows (cost matters)
- Simple execution where speed > quality
Use sonnet-general-purpose for:
- Multi-file reasoning and debugging
- Tasks requiring some judgment
- Daily coding work (80-90% of tasks)
Use opus-general-purpose for:
- Tasks requiring sustained focus and judgment
- When Sonnet keeps wandering or looping
- Complex analysis where staying on-track matters
- High-stakes decisions needing nuance
More from ed3dai/ed3d-plugins
functional-core-imperative-shell
Use when writing or refactoring code, before creating files - enforces separation of pure business logic (Functional Core) from side effects (Imperative Shell) using FCIS pattern with mandatory file classification
105playwright-debugging
Use when Playwright scripts fail, tests are flaky, selectors stop working, or timeouts occur - provides systematic debugging approach for browser automation issues
26researching-on-the-internet
Use when planning features and need current API docs, library patterns, or external knowledge; when testing hypotheses about technology choices or claims; when verifying assumptions before design decisions - gathers well-sourced, current information from the internet to inform technical decisions
22creating-an-agent
Use when creating specialized subagents for Claude Code plugins or the Task tool - covers description writing for auto-delegation, tool selection, prompt structure, and testing agents
18writing-for-a-technical-audience
Use when writing documentation, guides, API references, or technical content for developers - enforces clarity, conciseness, and authenticity while avoiding AI writing patterns that signal inauthenticity
18asking-clarifying-questions
Use after initial design context is gathered, before brainstorming - resolves contradictions in requirements, disambiguates terminology, clarifies scope boundaries, and verifies assumptions to prevent building the wrong solution
17