talk-roberts-ai-native-brownfield

Use when the user asks about Katie Roberts''s talk "Stop Maintaining, Start Evolving: Applying AI-Native Practices to Brownfield Codebases" — including questions about using AI to build large complex systems (her ~350k-line Rust S3 clone experiment), test oracles, flaky tests with AI agents, why 100% test coverage is the wrong goal, human-in-the-loop AI coding, AI-assisted performance engineering, using the type system to enforce invariants, tracing as an AI debugging tool, or applying her approach to brownfield/legacy modernisation work.

Quality

91%

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

82%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 well-structured reference skill body with concrete, actionable grounding workflow and clean one-level-deep file navigation. It loses points only for mild redundancy across the "How to help" subsections and a small inconsistency in where supporting files are listed.

Suggestions

Consolidate the repeated "Follow the workflow above..." references in the four "How to help" subsections into a single shared pointer to reduce redundancy.

List quote.md alongside outline.md and transcript.md in the "Supporting files" section so all referenced files are discoverable in one place.

Strengthen the workflow checkpoint by explicitly requiring excerpts to be confirmed verbatim against transcript.md before they are cited.

Dimension	Reasoning	Score
Conciseness	The body is mostly lean and avoids explaining concepts Claude already knows (it does not define test oracles or flaky tests), but the four "How to help" subsections each re-state "Follow the workflow above to retrieve and quote...", which could be consolidated.	4 / 5
Actionability	For an instruction-only reference skill it gives concrete, executable guidance: read outline.md, read specific transcript ranges, answer with safe excerpts, cite line numbers, use exact fallback phrasing ("the talk doesn't address this", "an audience member asked...").	5 / 5
Workflow Clarity	The five-step grounding workflow is clearly sequenced with an explicit verification checkpoint (step 4: if a claim is not in transcript.md, say so) and a speaker-attribution caution, but there is no feedback loop — acceptable for a Q&A skill yet short of the anchor-5 checklist/loop bar.	4 / 5
Progressive Disclosure	The body is a clear overview pointing one level deep to outline.md, transcript.md, and quote.md, each well signaled; however quote.md is introduced only in the later "Key quotes" section rather than listed alongside the others in "Supporting files", a minor organization gap.	4 / 5
	Total	17 / 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 excellent reference-skill description: it opens with an explicit "Use when..." trigger, lists concrete natural-language phrases users would say, and pins the skill to one distinct talk. The enumeration is long but each clause adds genuine trigger coverage rather than padding.

Dimension	Reasoning	Score
Specificity	The description enumerates many concrete topics from the talk ("test oracles", "flaky tests with AI agents", "100% test coverage", "type system to enforce invariants", "tracing as an AI debugging tool", "brownfield/legacy modernisation"), giving comprehensive coverage of what the reference skill addresses rather than vague language.	5 / 5
Completeness	It explicitly answers "when" ("Use when the user asks about...") with concrete trigger phrases and answers "what" by naming the talk, the speaker, the ~350k-line Rust S3 experiment, and the specific themes covered.	5 / 5
Trigger Term Quality	It leads with a natural "Use when the user asks about..." trigger and lists numerous phrases a user would actually say ("human-in-the-loop AI coding", "AI-assisted performance engineering", "flaky tests with AI agents", "100% test coverage"), with synonym variations like "brownfield/legacy modernisation".	5 / 5
Distinctiveness Conflict Risk	It targets a single named talk by a specific speaker with highly specific trigger themes, giving it a clear niche with minimal overlap risk against other skills.	5 / 5
	Total	20 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: jscraik/Agent-Skills
Path: Plugins/aidevcon/skills/talk-roberts-ai-native-brownfield/SKILL.md
Commit: a3808a9

Reviewed: ~3 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.