codesift-mcp 0.3.0 → 0.5.0

package/rules/gemini.md

@@ -10,15 +10,16 @@ For multi-repo sessions, call `list_repos()` to discover available repos.
 
 ## Tool Discovery
 
-**66 MCP tools total** (35 core visible + 31 discoverable).
+**146 MCP tools total** (51 core visible + 95 discoverable).
 
-35 core tools appear in ListTools. Remaining 31 niche tools are discovered on demand:
+51 core tools appear in ListTools. The remaining 95 niche tools are discovered on demand:
 
-- `discover_tools(query="dead code")` — keyword search across all 66 tools
+- `plan_turn(query="...")` — natural-language router with auto-reveal (preferred start-of-task entry point)
+- `discover_tools(query="dead code")` — keyword search across all 146 tools
 - `describe_tools(names=["find_dead_code"])` — get full parameter schema
 - `describe_tools(names=["find_dead_code"], reveal=true)` — also reveal in ListTools
 
-Core tools (35) always visible — includes search, symbols, context, analysis, architecture, conversations, and meta tools. Use `discover_tools` for niche tools like `classify_roles`, `find_unused_imports`, `rename_symbol`, `ast_query`, etc.
+Core tools (51) always visible — includes search, symbols, context, analysis, architecture, conversations, meta, and `plan_turn`. Use `discover_tools` for niche tools like `classify_roles`, `find_unused_imports`, `rename_symbol`, `ast_query`, etc.
 
 ## Tool Mapping
 
@@ -62,6 +63,56 @@ Use this table to pick the right tool for each task:
 | index conversations | `index_conversations` |
 | structural diff | `diff_outline(since=)` |
 | what changed | `changed_symbols(since=)` |
+| start of task / unclear where to begin | `plan_turn(query=...)` |
+| React component tree | `trace_component_tree` |
+| React hook analysis | `analyze_hooks` |
+| React re-render analysis | `analyze_renders` |
+| React context graph | `analyze_context_graph` |
+| find components | `search_symbols(kind="component")` |
+| find hooks | `search_symbols(kind="hook")` |
+| Astro islands/hydration | `astro_analyze_islands` |
+| Astro hydration audit | `astro_hydration_audit` |
+| Astro route map | `astro_route_map` |
+| Astro config analysis | `astro_config_analyze` |
+| Next.js full audit | `framework_audit` |
+| route map + rendering | `nextjs_route_map` |
+| SEO / metadata gaps | `nextjs_metadata_audit` |
+| server actions security | `framework_audit(checks=["server-actions"])` |
+| API contract extraction | `framework_audit(checks=["api-contract"])` |
+| client boundary / bundle | `framework_audit(checks=["boundary"])` |
+| broken internal links | `framework_audit(checks=["link-integrity"])` |
+| data fetch waterfalls | `framework_audit(checks=["data-flow"])` |
+| middleware coverage gaps | `framework_audit(checks=["middleware"])` |
+| server vs client classify | `framework_audit(checks=["components"])` |
+| Python project health audit | `python_audit` (compound: 8 checks, health score) |
+| Django/Flask/FastAPI routes | `trace_route(path="/api/users")` |
+| Python anti-patterns | `search_patterns("mutable-default"\|"bare-except"\|"eval-exec"\|"shell-true"\|"pickle-load"\|"n-plus-one-django")` |
+| async/await pitfalls | `analyze_async_correctness` (blocking requests/sleep, sync ORM in async) |
+| FastAPI Depends() graph | `trace_fastapi_depends` (Security, scopes, yield deps) |
+| Pydantic model schemas | `get_pydantic_models(output_format="mermaid")` |
+| Django/SQLAlchemy ORM graph | `get_model_graph` |
+| Django settings security audit | `analyze_django_settings` (15 checks) |
+| pytest fixture graph | `get_test_fixtures` (conftest hierarchy, scope, autouse) |
+| Celery task + canvas tracing | `python_audit(checks=["celery"])` |
+| Django signals / Celery / middleware | `find_framework_wiring` |
+| Python cross-module callers | `find_python_callers(target_name=)` |
+| Python circular imports | `python_audit(checks=["circular-imports"])` |
+| run Ruff on Python | `run_ruff(categories=["B","PERF","SIM","S"])` |
+| run mypy / pyright | `run_mypy` or `run_pyright` |
+| parse pyproject.toml | `parse_pyproject` |
+| Python dep freshness + CVEs | `analyze_python_deps(check_pypi=true, check_vulns=true)` |
+| Hono full overview | `analyze_hono_app` |
+| Hono middleware chain (route/scope/app-wide modes) | `trace_middleware_chain` |
+| Hono conditional middleware (applied_when) | `trace_middleware_chain(only_conditional=true)` |
+| Hono inline handler analysis | `analyze_inline_handler` |
+| Hono context variables | `trace_context_flow` |
+| Hono API contract / OpenAPI | `extract_api_contract` |
+| Hono response types (Issue #4270) | `extract_response_types` |
+| Hono RPC types / Issue #3869 | `trace_rpc_types` |
+| Hono security + env-regression audit / #3587 | `audit_hono_security` |
+| Hono architecture modules / #4121 | `detect_hono_modules` |
+| Hono dead routes | `find_dead_hono_routes` |
+| Hono route visualization | `visualize_hono_routes` |
 
 ## When to Use (Situational Triggers)
 
@@ -92,6 +143,21 @@ Use this table to pick the right tool for each task:
 | security audit | `scan_secrets(min_confidence="high")` |
 | code audit | `search_patterns("empty-catch")` |
 | past decisions | `find_conversations_for_symbol` |
+| starting a task, want tool recommendations | `plan_turn(query=...)` |
+| React codebase | `trace_component_tree("App")` + `analyze_hooks` |
+| React anti-patterns | `search_patterns("hook-in-condition")` |
+| React call graph (clean) | `trace_call_chain(filter_react_hooks=true)` |
+| new Next.js project (first look) | `framework_audit` |
+| SEO audit / metadata review | `nextjs_metadata_audit` |
+| security review (Next.js) | `framework_audit(checks=["server-actions"])` |
+| rendering strategy check | `nextjs_route_map` |
+| bundle size concerns | `framework_audit(checks=["boundary"])` |
+| new Hono project (first look) | `analyze_hono_app` |
+| Hono middleware order bug | `trace_middleware_chain` |
+| Hono auth false positive (blog pattern) | `trace_middleware_chain(only_conditional=true)` |
+| Hono handler introspection | `analyze_inline_handler` |
+| Hono RPC compile time slow | `trace_rpc_types` (flags Issue #3869 slow pattern) |
+| Hono enterprise architecture | `detect_hono_modules` (Issue #4121) |
 
 ## Key Parameters
 
@@ -99,7 +165,7 @@ Use this table to pick the right tool for each task:
 - `detail_level="compact"` — locations only (~15 tok/result vs ~150 default)
 - `token_budget=N` — cap output instead of guessing `top_k`
 - `file_pattern=` — always pass when scope is known (e.g. `"*.ts"`, `"src/tools/"`)
-- `kind=` — filter by type: `function`, `class`, `type`, `interface`
+- `kind=` — filter by type: `function`, `class`, `type`, `interface`, `component`, `hook`
 - `include_source=true` — include source code in results
 
 ### search_text