mcp-local

"Browse Bitbucket Cloud repositories and pull requests, read PR diffs, and create pull requests via the {{DISPLAY_NAME}}. Triggers on: list repos/branches/PRs, view a pull request or its diff, open/create a PR. PR creation is gated and supports dryRun."

Audit SQL construction to ensure parameterized execution, anti-concatenation posture, and read-only safety.

Control DB query budget with timeout, row limits, and concurrency caps to prevent abuse and resource exhaustion.

Validate MCP contract stability for tools/list, tools/call, schemas, and backward compatibility before merge or release.

Standardize MCP errors into user-actionable vs developer/internal classes with stable codes and safe messages.

Harden MCP host/client integration with token hygiene, scope minimization, transport safeguards, and policy checks.

Define operational telemetry and incident runbooks for MCP servers: logs, metrics, alerts, and recovery paths.

Run a pre-release checklist for local MCP packages: typecheck/build/start path, env constraints, README completeness, and safe defaults. Use before publishing or handing off.

Scaffold or extend local MCP servers in TypeScript/Node.js with zod schemas, clear tool contracts, and workspace conventions. Use when creating new MCP tools or new MCP server packages.

Review MCP server changes for security guardrails: least privilege, read-only DB defaults, SQL safety checks, secrets handling, and safe logging. Use before merging DB-related MCP changes.

Register a new MCP server in this workspace so it auto-gets an installer + native skill, and write a strong operational SKILL.md the AI will actually invoke. Use when adding a server to scripts/lib/manifest.mjs, authoring <server>/skill/SKILL.md, or improving an existing operational skill's trigger/guardrails.

Apply consistent read-only, idempotent, and destructive hints to MCP tools to improve client safety and UX.

Scaffold an internal MCP server for intelligent codebase indexing and graph queries. Use for creating index_repository/dependency/call-chain analysis capabilities backed by DB storage.

Design graph schema for code intelligence storage (dependencies, call chain, flow) with internal DB constraints. Use when modeling nodes/edges for codebase indexing MCP.

Implement hash-based incremental indexing for codebase intelligence MCP to avoid full re-index and keep graph data fresh efficiently.

Verify correctness parity between full and incremental indexing for the same revision.

Ensure index runs store complete provenance metadata for auditability and reproducibility.

Run pre-release checks for internal codebase indexing MCP: correctness, performance, security posture, and runbook readiness.

Review security for internal codebase-index MCP: data isolation, secret redaction, access control, safe logging, and controlled query exposure.

Define safe handling for unresolved symbols during extraction and graph linking without silent data loss.

Run mcp-local code exploration and impact analysis with MCP-first workflow, strict gates, and bounded fallback.

Extract symbols, imports, and call edges from source code using tree-sitter for codebase indexing MCP workflows. Use when implementing parser-based code intelligence.

"Use the {{DISPLAY_NAME}} for code structure analysis: index repos, search symbols, trace call chains, find change blast radius, and run safe rule-based refactors. MCP-first — query the graph before reading files. Triggers on: find symbol/function/class, who calls X, impact of changing X, rename/refactor, dead code, circular deps."

"Search logs and traces for the CommunicationHub backend via the {{DISPLAY_NAME}} (self-hosted OpenObserve). Triggers on: search logs, find errors/exceptions, trace a request, follow a trace id end-to-end, get spans, log volume/stats over a time window. Read-only."

"Query PostgreSQL safely via the {{DISPLAY_NAME}}: read-only SQL, table/schema inspection, multi-environment access, and gated writes/migrations. Triggers on: query the database, run SQL, inspect a table/schema, compare environments, diff data, EF Core migration. Read-only by default; prod is always read-only."