Text of github.github.com › gh-aw › reference › open-telemetry-attributes as Styx crawled it on Nov 19, 2026. The page may have changed since. Only its words are kept, without pictures or layout, and only the first part of a long page.

Open the live page

OpenTelemetry Attribute Reference | GitHub Agentic Workflows

Skip to main content OpenTelemetry Attribute Reference This page is the low-level attribute inventory for built-in OpenTelemetry emitted by GitHub Agentic Workflows. For setup and usage, see OpenTelemetry . Unless stated otherwise, attributes are emitted only when the underlying value is available. Different spans carry different subsets of these fields. Resource attributes Section titled “Resource attributes” Resource attributes identify the workflow run, repository, runner, and gh-aw environment attached to emitted spans. observability.otlp.resource-attributes can append custom resource attributes at runtime.

Use static strings or GitHub Actions expressions. Do not use secrets.* or vars.* values for this field because resource attributes are exported to external observability backends and are not treated as secret values. Attribute Description service.name OTel service name for the emitting component or workflow. service.version Version of the emitting gh-aw helper when available. gh-aw.workflow.name Workflow name. gh-aw.repository GitHub repository in owner/repo form. gh-aw.run.id GitHub Actions run ID. gh-aw.engine.id Raw gh-aw engine identifier when available. github.run_id GitHub run ID compatibility resource attribute. github.repository GitHub repository compatibility resource attribute. github.run_attempt GitHub Actions run attempt. github.event_name Triggering GitHub event name. github.ref Git ref for the run. github.ref_name Short Git ref name. github.head_ref Head ref for pull request-style events. github.sha Commit SHA. github.job GitHub Actions job ID. github.workflow_ref Workflow reference including path and ref. github.actor_id Numeric GitHub actor ID when available. runner.os Runner operating system. runner.arch Runner architecture. runner.name Runner name. runner.environment Runner environment classification. gh-aw.awf.version Workflow compiler version when available. gh-aw.awmg.version MCP gateway version when available. deployment.environment staging or production depending on staged mode.

Workflow and job span attributes Section titled “Workflow and job span attributes” These attributes appear on built-in workflow setup, agent, and conclusion spans as applicable. Attribute Description gh-aw.job.name Job name when available. gh-aw.workflow.name Workflow name. gh-aw.run.id Run ID. gh-aw.run.attempt Run attempt number. gh-aw.run.actor GitHub actor login. gh-aw.repository Repository in owner/repo form. gh-aw.cli.version CLI/helper version when available. gh-aw.event_name Triggering event name. gh-aw.deployment.state Deployment mode/state for setup or conclusion spans. gh-aw.workflow_run.conclusion Workflow run conclusion when available. gh-aw.staged Boolean staged-mode flag. gh-aw.trigger.item_type Trigger target type, for example issue or pull request. gh-aw.trigger.item_number Trigger target number. gh-aw.trigger.label Matched trigger label when label-based triggering is used. gh-aw.trigger.comment_id Comment ID for comment-driven triggers. gh-aw.frontmatter.source Frontmatter source classification. gh-aw.frontmatter.emoji Configured frontmatter emoji when present. gh-aw.frontmatter.body_modified Whether frontmatter processing modified the body. gh-aw.run.status Final workflow run status. gh-aw.error_count Count of output errors captured by the helper. gh-aw.warning_count Count of warnings observed during the run. gh-aw.permission_denied_count Permission denials observed during execution. gh-aw.steering_event_count Count of steering events recorded during the run. gh-aw.action_minutes Elapsed runtime converted to minutes. gh-aw.tracker.id Tracker identifier when present. gh-aw.aic AI credits consumed by the cost-owning job (agent, detection, or engine), encoded as a double when available.

Explicit zero is preserved; missing cost data is omitted, not zero-filled. gh-aw.max_ai_credits Configured max AI credits budget for the run when available. gh-aw.max_ai_credits_exceeded Whether the run exceeded the max AI credits budget. gh-aw.ai_credits_rate_limit_error Whether an AI-credits rate-limit or budget-exhaustion signal was detected. gh-aw.turns Total agent turns recorded for the run. gh-aw.integrity.filtered_events Integrity or secrecy filtering events recorded for the run.

Emitted on the conclusion span. gh-aw.firewall.requests Total firewall requests recorded for the run. Emitted on the conclusion span. gh-aw.firewall.allowed_requests Firewall requests allowed for the run. Emitted on the conclusion span. gh-aw.firewall.blocked_requests Firewall requests blocked for the run. Emitted on the conclusion span. gh-aw.mcp.tool_calls MCP tool calls recorded for the run. Emitted on the conclusion span. gh-aw.mcp.failed_tool_calls Failed MCP tool calls recorded for the run.

Emitted on the conclusion span. gh-aw.mcp.input_bytes Total serialized MCP tool input bytes. Emitted on the conclusion span. gh-aw.mcp.max_input_bytes Largest serialized MCP tool input payload in bytes. Emitted on the conclusion span. gh-aw.mcp.output_bytes Total serialized MCP tool output bytes. Emitted on the conclusion span. gh-aw.mcp.max_output_bytes Largest serialized MCP tool output payload in bytes. Emitted on the conclusion span. gh-aw.working_set.measurement_state Working-set measurement availability: measured , partial , or unavailable . gh-aw.working_set.rebuild_factor Cumulative canonical input tokens divided by peak invocation input tokens.

Omitted when unavailable. gh-aw.working_set.cumulative_input_tokens Sum of canonical logical input tokens across measured agent invocations. gh-aw.working_set.peak_input_tokens Largest canonical logical input-token count among measured agent invocations. gh-aw.working_set.rebuild_excess_tokens Cumulative input tokens minus peak input tokens, clamped to zero. gh-aw.working_set.invocations Valid invocation records contributing to the working-set measurement. gh-aw.agent.conclusion Normalized agent conclusion. gh-aw.detection.conclusion Detection subsystem conclusion when present. gh-aw.detection.reason Detection subsystem reason when present. gh-aw.otlp.export_errors Count of OTLP export errors seen during the run, including partial rejection and unknown acceptance due to invalid response diagnostics.

Counts export errors, not rejected records; warning-only partial success is excluded. gh-aw.otlp.export_error_details Serialized sanitized OTLP export error details when present, including signal-specific partial-rejection counts. Collector response messages are withheld. gh-aw.error.count Count of output errors on the conclusion span. gh-aw.error.messages Joined error messages when present. gh-aw.output.item_count Count of output items produced by the run. gh-aw.output.item_types Comma-separated output item types. gh-aw.github.rate_limit.remaining Last observed GitHub API rate limit remaining value. gh-aw.github.rate_limit.limit Last observed GitHub API rate limit ceiling. gh-aw.github.rate_limit.used Last observed GitHub API rate limit usage. gh-aw.github.rate_limit.resource GitHub rate limit resource bucket. gh-aw.github.rate_limit.reset GitHub rate limit reset timestamp.

For AIC rollups, use gh-aw.aic on the dedicated agent span ( gh-aw.agent.agent ) or, when no dedicated span exists, the cost-owning job’s conclusion span. It is emitted once per owning job, alongside gh-aw.workflow.name , gh-aw.run.id , and gh-aw.repository . Individual model/tool spans and downstream non-owning jobs need not carry AIC. Deduplicate spans mirrored to multiple backends by trace ID and span ID. Missing or unqueryable cost is unknown, not zero; backend field-validation errors alone do not prove an exporter failure.

Working-Set Rebuild Factor measures cumulative context reconstruction relative to peak invocation context. It is an efficiency/trajectory metric, not a measurement of semantic coherence debt and not a predictor of task success. Numeric working-set attributes are emitted on the built-in conclusion span only when the measurem

The rest of this page is on the live site.