<skill_overview> <purpose>Implement structured logging and tracing for Rust services</purpose> <triggers> <trigger>Adding logs or diagnostics</trigger> <trigger>Creating spans around operations</trigger> <trigger>Configuring tracing subscribers</trigger> <trigger>Ensuring sensitive data is not logged</trigger> </triggers> <sources> <source url="https://github.com/tokio-rs/tracing">Tracing GitHub</source> </sources> </skill_overview> <events_and_spans> <rules> <rule>Use spans for operations with duration</rule> <rule>Use events for point-in-time facts</rule> <rule>Attach structured fields to both spans and events</rule> </rules> </events_and_spans> <structured_logging> <rules> <rule>Prefer structured fields over string interpolation</rule> <rule>Use consistent field names and types</rule> </rules> <example> <code> tracing::info!(user_id, action = "login", "user action"); </code> </example> </structured_logging> <instrumentation> <rules> <rule>Use #[instrument] to create spans automatically</rule> <rule>Skip sensitive fields with #[instrument(skip(secret))]</rule> </rules> <example> <code> #[tracing::instrument(skip(password))] fn authenticate(username: &str, password: &str) { /* ... */ } </code> </example> </instrumentation> <subscriber_setup> <rules> <rule>Initialize a subscriber early in main</rule> <rule>Use env-based filtering for log levels</rule> <rule>Keep default level reasonable for production</rule> </rules> <example> <code> tracing_subscriber::fmt().with_env_filter("info").init(); </code> </example> </subscriber_setup> <sensitive_data> <never_log> <item>Passwords and tokens</item> <item>API keys and secrets</item> <item>PII when not required</item> </never_log> </sensitive_data> <anti_patterns> <avoid name="println_logging">Avoid println! for production logging</avoid> <avoid name="log_spam">Avoid high-cardinality fields in info logs</avoid> </anti_patterns>