Skip to main content
skill CX Codex

otel-ottl — Claude Code skill: what it does, how to install

Helps write and debug OTTL statements for OpenTelemetry Collector components. It assists with telemetry filtering, attribute redaction, and routing rules.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add dash0hq/agent-skills --skill otel-ottl -a claude-code
Open on GitHub

What otel-ottl does

The skill equips coding agents with syntax rules and context conventions for the OpenTelemetry Transformation Language (OTTL). It covers data path traversal across span, log, metric, datapoint, and resource contexts. OTTL strictly differentiates function types: converters such as ToUpperCase or Substring are uppercase and return values, while editors like set and delete_key are lowercase and modify fields in place.

It provides patterns for sanitizing sensitive fields, handling missing timestamps, and normalizing high-cardinality attributes. When testing whether an attribute exists, OTTL requires checking against nil rather than null. Syntax mistakes fail early during collector initialization. Always run otelcol validate --config=config.yaml before rolling out configurations.

Setting error_mode explicitly prevents silent telemetry loss in production. The transform and filter processors default to ignore in OpenTelemetry Collector Contrib v0.153.0 and newer, whereas connectors like routing propagate errors and drop the entire payload by default.

What it helps with

  • redact sensitive fields like database statements and authorization tokens
  • normalize high-cardinality attributes across span paths
  • route and filter telemetry payloads using conditional where clauses
  • configure the transform processor with safe error handling modes
  • validate path expressions across span, metric, and log contexts

Example prompts for the agent

  • Write an OTTL transform rule to redact db.statement when service.name is accounting
  • How do I drop spans using a filter processor where the path matches a regex?
  • Show me how to safely parse JSON from a log body attribute using OTTL

How to install otel-ottl

1Skills CLI (Vercel)

npx skills add dash0hq/agent-skills --skill otel-ottl -a claude-code

2Manually: put the files in place

~/.claude/skills/otel-ottl/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/otel-ottl/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about otel-ottl

How to install otel-ottl in Claude Code?

Run npx skills add dash0hq/agent-skills --skill otel-ottl -a claude-code. This downloads the skill definition directly into your agent environment.

What does the otel-ottl skill do?

It guides AI assistants when writing and debugging OTTL expressions for OpenTelemetry Collector processors and connectors. It ensures valid context usage, function syntax, and defensive error handling.

How does error handling differ between OTTL components?

Default behavior depends on the component. The transform and filter processors default to ignore starting in Contrib v0.153.0, but connectors like routing and signaltometrics default to propagate, which drops the payload on error.

Source text

SKILL.md · Apache-2.0
---
name: otel-ottl
description: OpenTelemetry Transformation Language (OTTL) expert. Use when writing or debugging OTTL expressions for any OpenTelemetry Collector component that supports OTTL (processors, connectors, receivers, exporters). Triggers on tasks involving telemetry transformation, filtering, attribute manipulation, data redaction, sampling policies, routing, or Collector configuration. Covers syntax, contexts, functions, error handling, and performance.
metadata:
  author: dash0
  version: '1.0.0'
---

# OpenTelemetry Transformation Language (OTTL)

## Components that use OTTL

OTTL is not limited to the transform and filter processors.
Processors (transform, filter, tailsampling, logdedup) and connectors (routing, count, sum, signaltometrics) accept OTTL expressions.
See [components](./rules/components.md) for the full list with use cases.

## OTTL syntax

### Path expressions

Navigate telemetry data using dot notation:

```
span.name
span.attributes["http.method"]
resource.attributes["service.name"]
```

**Contexts** (first path segment): `resource`, `scope`, `span`, `spanevent`, `metric`, `datapoint`, `log`.

### Enumerations

Use int64 constants for enumeration fields:

```
span.status.code == STATUS_CODE_ERROR
span.kind == SPAN_KIND_SERVER
```

### Operators

Assignment: `=` — Comparison: `==`, `!=`, `>`, `<`, `>=`, `<=` — Logical: `and`, `or`, `not`

### Functions

**Converters** (uppercase, return values):

```
ToUpperCase(span.attributes["http.request.method"])
Substring(log.body.string, 0, 1024)
Concat(["prefix", span.attributes["request.id"]], "-")
IsMatch(metric.name, "^k8s\\..*$")
```

**Editors** (lowercase, modify data in-place):

```
set(span.attributes["region"], "us-east-1")
delete_key(resource.attributes, "internal.key")
limit(log.attributes, 10, [])
```

See [function-reference](./rules/function-reference.md) for the full list of editors and converters.

### Conditional statements

Use `where` to apply transformations conditionally:

<!-- eval:skip -->
```
span.attributes["db.statement"] = "REDACTED" where resource.attributes["service.name"] == "accounting"
```

### Nil checks

Use `nil` for absence checking (not `null`):

```
resource.attributes["service.name"] != nil
```

## Validation workflow

1. **Validate config syntax** — run `otelcol validate --config=config.yaml` to catch compilation errors before starting the Collector.
2. **Test with the debug exporter** — route transformed telemetry to a `debug` exporter and inspect the output:

```yaml
exporters:
  debug:
    verbosity: detailed

service:
  pipelines:
    traces:
      receivers: [otlp]
      processors: [transform]
      exporters: [debug]   # swap in production exporter once validated
```

3. **Set `error_mode: ignore` in production** — see [Error handling](#error-handling).
4. **Promote to production exporters** — replace `debug` with the production exporter.

## Common patterns

<!-- keep-in-sync: each entry must match a ## heading in ./rules/patterns.md -->
- [Set attributes](./rules/patterns.md#set-attributes)
- [Drop telemetry by pattern](./rules/patterns.md#drop-telemetry-by-pattern)
- [Drop stale data](./rules/patterns.md#drop-stale-data)
- [Backfill missing timestamps](./rules/patterns.md#backfill-missing-timestamps)
- [Filter processor example](./rules/patterns.md#filter-processor-example)
- [Transform processor example](./rules/patterns.md#transform-processor-example)
- [Defensive nil checks](./rules/patterns.md#defensive-nil-checks)
- [Redact sensitive data](./rules/redaction.md) — strategies: replace, mask, hash, delete, and drop.
- [Normalize high-cardinality attributes](./rules/cardinality.md) — path segments, IP masking, and attribute count/length limits.
- [Enrich telemetry with static attributes](./rules/enrichment.md)

## Error handling

### Compilation errors

Occur during processor initialization and prevent Collector startup:
- Invalid syntax (missing quotes)
- Unknown functions
- Invalid path expressions
- Type mismatches

### Runtime errors

Occur during telemetry processing:
- Accessing non-existent attributes
- Type conversion failures
- Function execution errors

### Error mode configuration

Always set `error_mode` explicitly; the default differs by component.
The transform and filter processors default to `ignore` since Collector Contrib v0.153.0 and to `propagate` before it.
The routing and signaltometrics connectors default to `propagate`.
The count and sum connectors have no `error_mode` key and always propagate errors.

| Mode | Behavior | When to use |
|------|----------|-------------|
| `ignore` | Logs the error and continues to the next statement | **Production and general use** — the default for the transform and filter processors |
| `propagate` | Returns the error up the pipeline, dropping the payload from the Collector | Development and strict environments where you want to catch every error |
| `silent` | Ignores errors without logging | High-volume pipelines with known-safe transforms where error logs are noise |

```yaml
processors:
  transform:
    error_mode: ignore
    trace_statements:
      - set(span.attributes["parsed"], ParseJSON(span.attributes["json_body"]))
```

Statements are a flat list; the Collector infers the context (`span`, `metric`, `datapoint`, `log`, and so on) from the path prefixes.
Use the object form with an explicit `context:` only when a statement group mixes paths that cannot be inferred to a single context, or when you need a group-level condition or `error_mode`.

## Performance

Use `where` clauses to skip items early.

```
# BAD — runs replace_pattern on every span
replace_pattern(span.attributes["url.path"], "/\\d+", "/{id}")

# GOOD — skips spans that lack the attribute
replace_pattern(span.attributes["url.path"], "/\\d+", "/{id}") where span.attributes["url.path"] != nil
```

## References

- [OTTL Guide](https://www.dash0.com/guides/opentelemetry-transformation-language-ottl)
- [OTTL Specification](https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/pkg/ottl)
- [OTTL Functions Reference](https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/pkg/ottl/ottlfuncs)
- [OTTL Playground](https://ottl.run)

© dash0 · Apache-2.0 · dash0hq/agent-skills

Similar skills

Everything for DevOps and infrastructure →

From the same set3

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor