CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

ainativedev/aidevcon-2026-nyc

AI Native DevCon 2026 New York — practical agent skills grounded in conference talks

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

application-template.mdskills/talk-pratik-patel-refined-rag-data/references/

Version-aware RAG refinement and release

Use this worksheet for a bounded answer domain. It adapts Patel's process and leaves results unfilled until measured.

Intended coverage

Audience: Products and supported versions: Current-use questions: Migration or historical questions: Authoritative sources: Questions requiring clarification: Questions that should return insufficient evidence:

Write expected answers and source references before altering the corpus.

Refinement ledger

Source or chunkVersion and purposeActionReasonProvenance retained
Keep / remove / consolidate / enrich

Separate content extraction from stale-data handling. Preserve meaningful headings and code. Do not delete older material that supports a valid migration question.

Data checks

Metadata requirements: Duplicate policy: Current-version deprecation checks: Conflict handling: Curated FAQ evidence: Historical-content routing: Rejected-content retention:

A concise FAQ must remain traceable to source evidence. Record uncertainty rather than normalizing conflicting facts into one confident statement.

Evaluation matrix

QuestionCategoryMust includeMust excludeSupporting sourceCriticality
Current / stale / new / migration / absent

Chosen pass criteria: Critical failures that block release: Model and retrieval settings: Comparison baseline: Observed failures: Remaining coverage gap:

Do not adopt a threshold simply because the demonstration mentioned it. Keep test composition and denominator visible.

Release record

Input snapshot: Transformations and versions: Fixed output identifier: Evaluation evidence: Release decision: Rollback target: Trigger for re-evaluation:

Separate scratch data from the released dataset. Compare historical releases when answers regress.

Next improvement

Data problem supported by evidence: Smallest transformation: Regression question: Expected cost or quality effect: Actual outcome:

Neither a smaller corpus nor a correct sample answer is a universal guarantee. Do not upload or share source material beyond the user's authorization.

skills

talk-pratik-patel-refined-rag-data

tile.json