agentsclimarketplace
Sign in

Otel ottl

Skill ollygarden/opentelemetry-agent-skills/skills/otel-ottl

OpenTelemetry Transformation Language (OTTL) expert for writing and debugging telemetry transformations in the OpenTelemetry Collector. Use when authoring or reviewing `transform`, `filter`, `tail_sampling` processor configs or `routing` connector configs, debugging OTTL syntax or semantics, transforming traces, metrics, logs, or profiles, or converting data-processing requirements into OTTL statements.From its SKILL.md

Install
npx -y skills add ollygarden/opentelemetry-agent-skills --skill otel-ottl

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

5.6 KB, as published. Nobody here has run it

OpenTelemetry Transformation Language (OTTL)

OTTL transforms or selects telemetry inside Collector components. This skill is pinned to collector-contrib v0.158.0. Function, path, default, and feature-gate availability varies by release; when the user's version differs, verify against the matching upstream tag.

Workflow

  1. Choose the component. transform rewrites, filter drops, tail_sampling decides whether to retain traces, and routing sends telemetry to pipelines. A component controls its available contexts and functions.
  2. Choose the lowest usable context. Lower contexts can read their parents (for example, a span can read resource.attributes), but parents cannot read children. Use datapoint for point attributes instead of traversing metric.data_points.
  3. Verify every emitted function, then write the statement. Confirm each function's exact identifier and signature in references/functions.md. If an identifier is absent, treat it as unsupported instead of deriving or substituting a plausible name. An editor such as set or delete_key mutates data and may have a where condition. Converters such as ParseJSON and IsMatch return values; they do not mutate.
  4. Set error behavior deliberately. ignore logs statement errors and continues; silent continues without logging; propagate returns the error and can cause the component to drop the payload. In v0.158, transform and filter default to ignore; routing also defaults to ignore while its beta default-error feature gate is enabled. For routing, ignore sends an errored payload to default_pipelines; configure that fallback or the payload is dropped.
  5. Verify end to end. Validate the exact Collector version, then send known telemetry and inspect file-exporter output. Use the telemetrygen recipe.
set(span.attributes["env"], "prod") where resource.attributes["env"] == nil

Load only what the task needs

  • Contexts — exact paths, hierarchy, enums, and request metadata.
  • Functions — editor/converter signatures and release availability.
  • Quick reference — component YAML, recipes, escaping, troubleshooting, and safe skeletons.

For a single path or function, read only the relevant section instead of loading the full catalogs.

Safety and correctness gates

  • Guard optional or polymorphic input before conversion: where x != nil, IsString(x), or the appropriate type check.
  • For JSON-object-only work, guard both the type and shape before calling ParseJSON, for example IsString(log.body) and IsMatch(log.body.string, "(?s)^\\s*\\{.*\\}\\s*$"). The RE2 (?s) flag admits pretty-printed objects containing newlines. Checking IsMap after parsing does not prevent arrays or scalar JSON from being parsed.
  • On a version-pinned request, confirm every chosen path and function against that release tag; do not assume a function listed for this skill's v0.158 anchor exists in an older release. For v0.156 JSON-object parsing, ParseJSON, IsString, and IsMatch are available without the v0.157 alpha lambda feature gate.
  • Request metadata is read-only and may contain credentials. Copy only explicitly allowlisted, non-sensitive keys. OTLP metadata routing requires include_metadata: true on the receiver. HTTP/client header spelling may retain its form (otelcol.client.metadata["X-Tenant"][0]); gRPC metadata keys are lowercase (otelcol.grpc.metadata["x-tenant"][0]).
  • The routing request context is deprecated as of v0.156; use otelcol.client.metadata or otelcol.grpc.metadata.
  • Log-record-specific rewrites of shared resource or scope data require flatten_data: true and the alpha transform.flatten.logs gate. This copies and regroups data; do not enable it accidentally.
  • Hashing an identifier does not necessarily anonymize it. Apply the organization's data-handling policy before retaining deterministic hashes of personal data.

Frequent syntax traps

  • In Collector YAML, write an OTTL replacement backreference ${1} as $${1}. A replacement such as $1REDACTED is literal and silently fails to substitute the capture.
  • Go RE2 rejects large counted repetitions such as (.{1024}).*; use Substring with a nil/type guard and Len, or truncate_all for a map.
  • Current span-event paths use spanevent.*, not span_event.*. Cache paths are context-qualified, such as span.cache["parsed"].
  • Use Decode(value, "base64"); Base64Decode is deprecated.
  • Regex escapes inside OTTL strings are doubled (\\d, \\s, \\.).

Upstream sources

What ships with it: 4 files

44.5 KB alongside SKILL.md

evals/

references/

Gives 0 of the 12 instructions most monitoring observability skills give

Counted across 530 of the 532 authors here whose files we hold, read 2026-09-06

  • Use structured JSON loggingin 40 of 530, across 36 files
  • Link every alert to a runbookin 29 of 530, across 27 files
  • Attach correlation IDs to every log linein 19 of 530, across 16 files
  • Alert on symptoms rather than causesin 19 of 530, across 17 files
  • Use OpenTelemetry for distributed tracingin 15 of 530, across 14 files
  • Alert on symptoms users feelin 15 of 530, across 13 files
  • Implement health check endpointsin 14 of 530, across 10 files
  • Inspect existing dashboards firstin 12 of 530, across 4 files
  • Build the minimum useful boardin 12 of 530, across 4 files
  • Start from operator questionsin 12 of 530, across 4 files
  • Propagate trace context across boundariesin 11 of 530, across 10 files
  • Include trace id in all log entriesin 10 of 530, across 9 files

Said here and by no other author read

  • Choose the component
  • Choose the lowest usable context
  • Verify every emitted function
  • Set error behavior deliberately
  • Verify end to end
  • Guard optional input before conversion

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.

Keep looking

Skills are one crate of 626,194. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.