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-knox-harness-engineering/references/

Factory improvement loop proposal

This is a recommended design artifact based on Knox's talk, not an automation to execute.

Problem and diagnostic baseline

Observed signal: Choose cost per PR, review rounds, merge time, volume or a directly observed failure. Record the actual baseline, comparison period and task mix. Mark missing measurements.

Supporting examples: Include representative trace excerpts, comments or CI failures. Distinguish the observed symptom from the suspected cause.

Category: Agent access/tooling, architecture, repetition, test quality or CI capacity. Explain why that category fits better than the alternatives.

Narrow loop

Input: Name the evidence the loop can read and its existing access boundary.

Analysis: State the pattern it should identify. For example, repeated malformed calls to the same issue-tracker operation may suggest a small validated wrapper rather than more prose.

Output: Choose a recommendation, issue or reviewable change. Define the intended consumer and preserve the existing merge policy.

Cadence and budget: Select a schedule appropriate to the evidence rate and cost. These are proposed settings, not numbers mandated by the talk.

Verification

Describe one representative fresh task and the failure the change should prevent. Record outcome, retries, relevant cost and verification quality. If a test or CI change is proposed, explain what meaningful coverage remains and what moves to another execution bucket.

For architectural signals, inspect before splitting: a large file or weakly connected cluster is a reason to investigate, not automatic proof of bad design.

Decision and stopping condition

State when to keep, revise or stop the loop, including repeated failures, noisy suggestions or rising cost without better outcomes.

Next action: Identify one concrete change and the evidence still required. Keep planned evaluations separate from completed results; do not claim improvement until the task has been observed after the change.

skills

talk-knox-harness-engineering

tile.json