CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

talk-roberts-brownfield-ai-native

Use when the user asks about Katie Roberts's talk "Stop Maintaining, Start Evolving: Applying AI-Native Engineering in Brownfield Codebases" (AI Native DevCon, June 2026) — including questions about brownfield vs greenfield AI engineering, the three methodologies (pseudo-greenfield, strangler fig pattern, branch by abstraction), the "code as a city" metaphor, using AI to map and modernize legacy codebases, planning skills and developer skills, the value-vs-complexity mirror exercise, avoiding AI agents going rogue on legacy code, the AG Grid upgrade case study, Nearform's "six months in eight weeks" pseudo-greenfield case study, or applying her brownfield AI-native approach to current legacy modernization work.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

85%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 operational guide: task-typed workflows with concrete steps, explicit grounding/validation checkpoints, and careful safety handling of untrusted transcript material and unreliable speaker attribution. The main gaps are the referenced bundle files being absent from the skill directory and some repetition of the grounding pattern across sections.

Suggestions

Ship the referenced bundle files (outline.md, transcript.md, quote.md) — e.g., in a references/ directory with the body's paths updated — so each workflow's first step is executable as shipped.

Consolidate the repeated 'read outline.md then transcript.md' and quote-safety steps into the top-level grounding rules, keeping per-section steps only where they add task-specific direction.

DimensionReasoningScore

Conciseness

The body is operational and dense with no padding or explanations of concepts Claude already knows, but the read-outline-then-transcript pattern and 'quote short, non-sensitive excerpts' phrasing are repeated across nearly every section and could be consolidated into the top-level grounding rules. Efficient with minor trims possible, short of the every-token-earns-its-place anchor.

4 / 5

Actionability

Every workflow gives concrete, executable direction: exact files to read, which outline sections to use ('Named frameworks / concepts', 'Terminology glossary'), named quote candidates, explicit verdict categories ('covered / partial / missing'), a specified multi-agent flow, and a placeholder marker '[not from talk — added as a starting placeholder]'. As an instruction-only skill this fully meets the concrete-guidance standard.

5 / 5

Workflow Clarity

Each task type has a numbered sequence with explicit checkpoints and feedback loops for grounding errors: 'If a claim isn't in transcript.md, say "the talk doesn't address this"', 'If the framework genuinely doesn't fit, say so', citation-by-line-range requirements, and the anti-fabrication attribution rules. This is a non-destructive Q&A skill, so the destructive/batch validation cap does not apply.

5 / 5

Progressive Disclosure

The body is a well-organized overview pointing one level deep to clearly signaled references (outline.md, transcript.md, quote.md) — but none of these files exist in the skill bundle (no references/ directory is present), so every workflow's first step cannot actually be completed as shipped. Good structure undermined by missing files.

3 / 5

Total

17

/

20

Passed

Description

72%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.

A rich, highly distinct trigger list that would reliably fire for anything touching this specific talk, with excellent natural keyword coverage. Its main weakness is that it never states what the skill does — the capability is only implicit in the 'Use when' enumeration.

Suggestions

Prepend an explicit capability statement (e.g., 'Answers questions about ... by grounding responses in the verified transcript and outline, quoting safe excerpts') before the 'Use when' clause so the 'what' is stated, not implied.

Trim the enumerated sub-triggers to distinct trigger classes — minor items like the mirror exercise and individual case-study names could fold into shorter phrases — to cut token cost without losing firing precision.

DimensionReasoningScore

Specificity

The domain is named in exhaustive concrete detail ('pseudo-greenfield, strangler fig pattern, branch by abstraction', 'AG Grid upgrade case study', 'value-vs-complexity mirror exercise'), but no explicit actions are stated — capabilities are only implicit in 'questions about' and 'applying her brownfield AI-native approach to current legacy modernization work'. This sits between anchor 2 (domain named, actions minimal/generic) and anchor 4 (several specific actions listed).

3 / 5

Completeness

The 'when' is exceptionally explicit ('Use when the user asks about Katie Roberts's talk...'), but the 'what' is never stated — the description does not say the skill answers questions grounded in a transcript/outline. The detailed trigger enumeration conveys real scope above the bare when-only anchor 2, but without an explicit capability statement it does not reach anchor 4.

3 / 5

Trigger Term Quality

Comprehensive natural-language coverage: 'Katie Roberts', the full talk title, 'brownfield vs greenfield AI engineering', 'legacy codebases', 'modernize legacy', 'strangler fig pattern', 'AI agents going rogue', plus both case-study names. No common variation or synonym a user would naturally say is missing.

5 / 5

Distinctiveness Conflict Risk

Pinned to a single named talk, speaker, and conference ('AI Native DevCon, June 2026') with distinctive topic triggers; it would essentially never fire for the wrong skill.

5 / 5

Total

16

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository

jscraik/Agent-Skills

Path

Plugins/aidevcon/skills/talk-roberts-brownfield-ai-native/SKILL.md

Reviewed

Table of Contents

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.