code-search

Pick the tool by the shape of the question:

• Exact identifier known → mcp__plugin_ora_fff__grep. Naming variants of one identifier (e.g. ['ActorAuth', 'actor_auth']) → mcp__plugin_ora_fff__multi_grep.
• File by name → mcp__plugin_ora_fff__find_files (frecency-ranked, git-dirty boosted).
• Concept / "how does X work" / "where is Y handled" (no identifier yet) → mcp__plugin_ora_morph__codebase_search. One call replaces the grep→Read→grep loop; fall back to fff + Read only when it misses.
• Symbol definition / all references → LSP.
• Structural patterns (match by AST, not text) → ast-grep skill.
• Map a file's or directory's symbols/exports/imports before reading → ast-grep outline (ast-grep skill).
• Git history / blame → git via Bash.

Morph returns a synthesized answer with file:line citations. The citations are real; the synthesis can misread what the code does — Read the cited locations before relaying load-bearing conclusions.

Don't shotgun-grep a feature by OR-ing guessed names (grep "FreeGift|GiftModal|BuyXGetY|percentOff") — synonyms get missed and output drowns in noise. Learn the feature's actual name from README or directory structure, grep one specific term, and follow references.