Instrument Rust applications with tracing spans, events, subscribers, and structured logging.
npx skills add https://github.com/avbel/ai-skills --skill tracing-rust-avbel
Please help me install this Agent Skill. Skill: tracing-rust Source: https://github.com/avbel/ai-skills/tree/main/skills/tracing-rust Command: npx skills add https://github.com/avbel/ai-skills --skill tracing-rust-avbel
#[instrument] attribute, manual span creation, field sigils (% and ?), and async instrumentation with .instrument(). - Subscriber Configuration: Shows fmt subscriber setup, EnvFilter with RUST_LOG, JSON output, layered registries, and tokio-console integration. - Structured Error Logging: Enforces structured error fields (error = ?e) instead of stringified messages, plus production guidance like release_max_level_info. - Use Case: When adding observability to an axum or tokio service, use this Skill to instrument handlers with #[instrument], configure an env-filtered JSON subscriber, and avoid common async pitfalls. ## Quick Start Instrument my Rust async handlers with tracing spans and set up a JSON subscriber filtered by the RUST_LOG environment variable.High-intent search queries and answers about installing and using this skill.
Add the #[instrument] attribute from the tracing crate to automatically create a span from the function name and arguments. Use skip() for large or sensitive parameters, and never call span.enter() across .await points—use .instrument() on futures instead.
Build an EnvFilter with EnvFilter::try_from_default_env() and pass it to tracing_subscriber::fmt().with_env_filter(filter).init(). This reads the RUST_LOG variable, supporting per-module directives like my_app=debug,sqlx=warn.
The % sigil formats fields with Display, suited for strings and numbers, while ? uses Debug, suited for structs, enums, and errors. No sigil defaults to Display for numbers and Debug for other types.
Yes, add console-subscriber as a dependency and call console_subscriber::init() in main. Enable tracing's release_max_level_info feature in production builds to limit overhead while keeping tokio-console support during development.
Tracing produces no output without an initialized subscriber. Call tracing_subscriber::fmt::init() early in main; without it, all spans and events are no-ops. Also verify RUST_LOG is not filtering out your target module.
Avoid high-cardinality fields like per-request IDs in span-level fields within hot loops, since each unique value creates a new span identity and increases overhead. Record such values on events instead of spans.
fusengineGuides crate selection for Rust projects across serialization, async, web, database, and error handling domains.
mariotoffiaStandardize logs, metrics, and distributed tracing across microservices.
NackalalalongUnify logs, traces, and metrics in Rust applications with tracing and Prometheus.
BagelHoleInstrument applications and infrastructure with OpenTelemetry for unified traces, metrics, and logs.