rerun-chunk-processing

Core mechanics of the Rerun Chunk Processing API (rerun.experimental) — LazyChunkStream pipelines, Chunk, lenses (MutateLens/DeriveLens/Selector), RrdReader, writing optimized RRDs. Read BEFORE writing any ingestion/conversion/preprocessing code (convert an MCAP, build a recording from a dataset, preprocess an .rrd, port an old converter): it mandates reader+lens pipelines and steers away from hand-built chunks — no Chunk.from_columns for data a reader/lens can produce, no per-message rr.log, no manual pa.array assembly. Source-specific knowledge lives in the importer skills (rerun-mcap, rerun-urdf, rerun-parquet, rerun-mp4, rerun-lerobot); read rerun-data-model first to decide what the data should become.

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

76%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A strong, actionable reference for the rerun.experimental chunk-processing API with executable examples and a clear decision tree, weakened only by some reinforcement repetition across sections and the absence of an explicit output-validation checkpoint in its batch ingestion workflow.

Suggestions

Add an explicit verification step to the conversion workflow (e.g., after write_rrd, read back with RrdReader(path) and check recording_id/row counts) to turn the batch ingestion into a validate→fix→retry loop and lift workflow_clarity above the batch cap.

Consolidate the repeated reader+lens-vs-hand-built guidance: the 'Anti-patterns', 'Decision rule', and 'Hand-built chunks — sidecar only' sections overlap — fold the sidecar list into the decision rule and keep one anti-pattern table to trim tokens.

Move the Selector grammar and lens output-mode reference into a references/ file (e.g. SELECTOR.md) and link to it, so SKILL.md stays a lean overview and progressive_disclosure can reach 5.

Dimension	Reasoning	Score
Conciseness	Dense and API-specific (assumes Claude knows general programming, explains only the niche rerun.experimental surface), with executable snippets earning their tokens; not a 5 because the 'Anti-patterns', 'Decision rule', and 'Hand-built chunks — sidecar only' sections reinforce the same reader+lens-vs-hand-built message with some repetition that could be trimmed.	4 / 5
Actionability	Provides copy-paste-ready executable code for the common cases (Chunk.from_columns sidecar, MutateLens in-place fix, DeriveLens unit conversion, merge→collect→write_rrd) with real imports and a concrete introspection command, matching the 5 anchor's fully-executable common-case coverage.	5 / 5
Workflow Clarity	The 'Decision rule' gives a clear 6-step sequenced procedure with anti-pattern guardrails and a gotchas section, which would be a 4, but the skill drives batch ingestion/conversion (whole datasets → RRD) with no explicit validate-the-output checkpoint or feedback loop, so the batch-operations cap holds it at 3.	3 / 5
Progressive Disclosure	Well-organized into clear headers (Core model, Stream composition, Lenses, Selector grammar, Writing RRDs, etc.) with one-level-deep external references (GitHub example, docs URLs) and a table routing to importer skills; not a 5 because all API-reference detail (Selector grammar, lens output modes) is inline in a single long file rather than split into reference files, and no bundle files exist to offload it.	4 / 5
	Total	16 / 20 Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exemplary description: third-person voice, concrete API primitives and tasks, explicit 'Read BEFORE...' trigger guidance with natural phrases and file extensions, and clear routing that distinguishes it from sibling importer skills. No fluff or over-claims despite its density.

Dimension	Reasoning	Score
Specificity	Names concrete API primitives ("LazyChunkStream pipelines, Chunk, lenses (MutateLens/DeriveLens/Selector), RrdReader, writing optimized RRDs") plus concrete tasks ("convert an MCAP, build a recording from a dataset, preprocess an .rrd, port an old converter") and explicit anti-patterns ("no Chunk.from_columns... no per-message rr.log, no manual pa.array assembly") — comprehensive concrete coverage, matching the 5 anchor rather than the 4 (which allows minor gaps).	5 / 5
Completeness	Explicitly answers 'what' (core mechanics of the Chunk Processing API) and 'when' ("Read BEFORE writing any ingestion/conversion/preprocessing code (convert an MCAP, build a recording from a dataset, preprocess an .rrd, port an old converter)") with concrete trigger phrases, satisfying the 5 anchor and not the 3-cap (an explicit trigger clause is present).	5 / 5
Trigger Term Quality	Includes natural user phrases ("convert an MCAP", "build a recording from a dataset", "preprocess an .rrd", "port an old converter") plus synonyms (convert/port, ingestion/conversion/preprocessing) and file extensions (MCAP, .rrd), matching the 5 anchor's comprehensive synonym+extension coverage rather than the 4 anchor's 'a few natural terms missing'.	5 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (Rerun rerun.experimental chunk processing) and explicitly routes source-specific work to named importer skills (rerun-mcap, rerun-urdf, etc.) and rerun-data-model, giving it distinct triggers with minimal conflict risk, matching the 5 anchor.	5 / 5
	Total	20 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: rerun-io/rerun
Path: skills/rerun-chunk-processing/SKILL.md
Commit: f30766b

Reviewed: ~8 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.