Manage the custom agent team (.agent.md files). Use when: creating, updating, reviewing, or deleting agent definitions under .github/agents/. Covers model assignment rules, frontmatter conventions, tool selection, and team structure.
MangoHud CSV log data column specifications — all 16 data columns with C++ types, output formats, units, value ranges, data sources, and update frequencies. Use when: implementing MangoHud CSV parsers, validating column data types, understanding what each metric measures, debugging unexpected values (zeros, duplicates, spikes), or mapping MangoHud columns to internal data structures.
MangoHud CSV log file format specification — file structure, system info header, data columns (FPS, frametime, CPU/GPU metrics), elapsed time, log_versioning variant, summary files, file naming conventions, and configuration options for generating logs. Use when: parsing MangoHud CSV files, writing MangoHud CSV exporters, validating benchmark upload data, understanding column data types/units/ranges, or debugging MangoHud log parsing issues.
MangoHud log file generation — configuration options (output_folder, log_interval, log_duration, autostart_log), file naming conventions, keybinds for toggling logging, summary file format and statistics, log_versioning option, upload to flightlessmango.com, and benchmark percentile configuration. Use when: configuring MangoHud for benchmark logging, understanding log timing/intervals, parsing summary CSV files, or troubleshooting why logs aren't generated.
MCP 2025-11-25 client features — sampling (createMessage, model preferences, hints, priorities, tool use in sampling, multi-turn tool loops, toolChoice modes, content constraints), roots (filesystem boundaries, list, notifications), elicitation (form mode with JSON Schema, URL mode for sensitive data, response actions, completion notifications, URLElicitationRequiredError). Use when: implementing MCP client capabilities, handling sampling requests, managing roots, processing elicitation.
Model Context Protocol (MCP) 2025-11-25 core specification — architecture (host/client/server), JSON-RPC 2.0 base protocol, lifecycle (initialization, capability negotiation, shutdown), error codes, _meta fields, icons, utilities (ping, cancellation, progress, tasks, pagination). Use when: implementing MCP servers or clients, designing MCP protocol messages, debugging MCP lifecycle issues, understanding MCP architecture.
MCP 2025-11-25 server features — resources (list, read, templates, subscriptions, annotations), tools (list, call, inputSchema, outputSchema, annotations, structured content, tool names), prompts (list, get, arguments, messages), completion (autocomplete for prompt/resource arguments), logging (setLevel, notifications, syslog levels). Use when: implementing MCP server capabilities, adding tools/resources/prompts, handling tool calls, resource subscriptions.
MCP 2025-11-25 transports and authorization — stdio transport (subprocess model, stdin/stdout JSON-RPC, stderr logging, shutdown sequence), Streamable HTTP transport (single endpoint, POST/GET/DELETE, SSE streaming, session management with MCP-Session-Id, resumability via Last-Event-ID, multiple connections, backwards compatibility), OAuth 2.1 authorization (Protected Resource Metadata RFC 9728, AS discovery, client registration approaches, PKCE S256, resource parameter RFC 8707, scope challenge handling, security constraints). Use when: implementing MCP transports, configuring HTTP endpoints, implementing OAuth for MCP, handling session management.