zoom-out

Summary

A meta-skill that instructs the agent to shift to a higher level of abstraction when analyzing unfamiliar code. Rather than diving deep, it requests a broad map of relevant modules, callers, and relationships using domain-specific vocabulary. This is a lightweight instruction skill with no executable operations.

Detected Capabilities

Trigger Keywords

Phrases that MCP clients use to match this skill to user intent.

Use Cases

• Understand how an unfamiliar code section fits into the larger system
• Get a high-level architecture overview when lost in implementation details
• Map module dependencies and caller relationships for onboarding
• Establish shared vocabulary and domain concepts before deep dives

Quality Notes

• Skill is very minimal — only 2 sentences of instruction with no elaboration or examples
• No guidance on what 'domain glossary vocabulary' means or how to extract it
• No examples of expected output format (e.g., dependency graph, module list, documentation structure)
• No edge case handling documented (e.g., monorepo projects, circular dependencies, undocumented modules)
• No limitations stated — unclear if this applies to all codebases or specific types
• The phrase 'go up a layer of abstraction' is vague — could mean different things depending on project structure
• Missing context: does this skill expect the agent to already have project context, or should it scan the project first?
• disable-model-invocation: true suggests this is a prompt-style instruction, not an operational skill, but this design choice is not explained
• Positive: the meta-skill approach is conceptually sound — shifting perspective without triggering a full model restart is useful