Skill v1.0.3
currentAutomated scan100/100+6 new, ~1 modified
version: "1.0.3" name: intuitive-init description: Create or slim project-local agent guidance and startup configuration from current repository evidence. disable-model-invocation: true
Intuitive Init
Create, audit, refresh, or slim project-local agent guidance from current repo evidence. Preserve commands, hazards, permission boundaries, and ownership rules. Generated init output is a suggestion to merge, never an overwrite source.
For ordinary guidance edits, inspect the target file and the commands/docs it references; a repo-wide scan, nested agent, or full manual read is optional. Keep shared rules in AGENTS.md and Claude-only additions in CLAUDE.md. Use $intuitive-doc for broad human-documentation cleanup.
Read for the selected task
| Task | Detail to read when needed | |
|---|---|---|
| Guidance health audit | Audit criteria | |
| Initial creation or refresh | Apply, Refresh, and merge rules | |
| Startup-context pressure | Startup hygiene and cleanup | |
| Long root instructions | Slim / cleanup | |
| Root file linked to a shared toolkit | Symlink migration | |
| Native init suggestions are useful | Init discovery | |
| LSP, Serena, language server, or MCP setup | Tool setup before editing configuration |
Use the indicated sections, not the entire detailed guide. Extract long procedures only when they have a real consumer, and leave a task-specific pointer at the root. Keep credentials and machine-local paths out of committed guidance.
When durable Intuitive workflow adoption is in scope, preserve the repo's planning/status paths and explicit project-integrator ownership. Do not create STATUS.md, a validator, or a status directory merely because Init ran.
Report changed files, verification, remaining risks, and whether generated init suggestions were merged, rejected, or not used. Stop when the accepted guidance scope is current, concise, and linked to the relevant detail.