Help us improve
Share bugs, ideas, or general feedback.
From rldyour-serena-mcp
Семантический воркфлоу для кода через Serena MCP - приоритет перед raw reads и grep. Используй для: изучи код, проиндексируй проект, найди символы, найди ссылки на символ, рефакторинг, проследи реализацию, инспектируй файл. EN triggers: navigate code, find symbol, find references, semantic refactor, trace usages, inspect file, code lookup, symbol search, where is X defined.
npx claudepluginhub nddev-it-com/rldyour-claudecode --plugin rldyour-serena-mcpHow this skill is triggered — by the user, by Claude, or both
Slash command
/rldyour-serena-mcp:serena-code-workflowThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use Serena MCP as the primary semantic layer for code work. The goal is low semantic entropy: understand structure before reading bodies, edit symbols instead of brittle line ranges, keep project knowledge synchronized, and prefer scalable patterns over quick text surgery.
Guides technical evaluation of code review feedback: read fully, restate for understanding, verify against codebase, respond with reasoning or pushback before implementing.
Share bugs, ideas, or general feedback.
Use Serena MCP as the primary semantic layer for code work. The goal is low semantic entropy: understand structure before reading bodies, edit symbols instead of brittle line ranges, keep project knowledge synchronized, and prefer scalable patterns over quick text surgery.
User-facing conversation stays in Russian unless the user asks otherwise. Repository documentation, code comments, commit messages, memory files, plan files, and research archive files are written in English.
Use this skill without waiting for explicit invocation when the task asks to:
If Serena is unavailable or the file type is not supported, state the fallback and continue with Claude Code-native tools (Read, Grep, Glob, Edit). Do not block progress just because a preferred MCP tool is missing.
For reading code, follow this order unless the task is trivial or the current session lacks a listed tool:
activate_project when the active project is unclear.check_onboarding_performed to verify whether Serena project onboarding exists.list_memories to discover durable project knowledge.read_memory only for relevant memories inferred from names.get_symbols_overview to map top-level file structure without reading full files.find_symbol with include_body=false to discover children, overloads, and symbol paths.find_symbol with include_body=true only for the exact implementation needed.find_referencing_symbols to trace callers, usages, dead code risk, and refactor impact.search_for_pattern only for broad text sweeps, non-symbol text, or fallback coverage.For editing code, prefer:
find_symbol(include_body=true) to read the current implementation.replace_symbol_body for whole-symbol replacement.insert_before_symbol or insert_after_symbol for structured additions.rename_symbol for LSP-aware symbol renames.safe_delete_symbol if available and suitable.Use raw rg, direct file reads, and line patches only when Serena cannot answer the question, when the task is text-level rather than symbol-level, or when the needed edit is a tiny known-location change.
For Serena project setup and indexing, follow Serena's project workflow. The official indexing operation is the Serena project command serena project index; MCP tools then use the resulting project and language-server state for semantic navigation. Do not invent a nonexistent MCP "index directory" tool.
When a project is not onboarded, use Serena onboarding patterns and memory tools as the source of durable project context. Full project initialization commands belong to the rldyour-flow plugin; this skill only defines the Serena-first rule and the correct tool priority.
Code is the source of truth. Memories, plans, and docs must be checked against actual code, git diff, and task-specific verification.
Do not read entire files when get_symbols_overview plus targeted find_symbol is enough. Do not use grep as the first move for symbol-level questions.
For non-trivial decisions, make a short explicit plan before editing. If the task requires external technical research, use the rldyour-explore plugin first and then return to Serena for local code impact.
After meaningful project changes, use the serena-memory-sync skill before final stop so .serena/memories stays accurate.
For investigation-only work, report:
Scope: what code was inspected.Evidence: Serena tools or fallback sources used.Findings: concrete facts with file paths or symbol names.Next steps: implementation or verification actions.For implementation work, keep the final concise: what changed, how it was verified, and any residual risk.
get_symbols_overview for symbol searchreplace_symbol_body would do