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-pari-patel-agent-awareness/references/

User-owned interaction policy

This is a recommended application of Patel's talk. It does not infer medical or psychological diagnoses.

Current task and explicit preferences

Record whether the user is exploring, implementing, reviewing or handling an urgent issue. Include only context actually supplied or legitimately available.

Separate explicit preferences from hypotheses based on interaction. Do not turn a long session, delayed reply or short answer into a claim about fatigue or competence.

Visible state

FieldCurrent valueUser-set or inferredEvidence and review point
Explanation depthFill from taskMark sourceInclude uncertainty
Interrupt budgetUser-controlledMark sourceDefine exceptions
Decision scopeExisting authorizationExplicitNever expand by inference
Task modeCurrent workMark sourceRevisit when task changes

Confidence and review timing are recommended additions to the talk's four fields.

Behavior rules

Depth: State what a concise answer includes and where optional detail appears.

Pacing: Define cooldown, batching and urgent-blocker behavior. Continue independent work while a required decision is pending.

Handoff: Provide findings, options, their consequences, a recommended default and when an answer is needed. Defaults apply only within already-delegated scope; silence is not approval.

Override: Explain how the user can see and change the state immediately.

Evaluation

Track re-asks, decision delay, overrides and the two distinct interruption complaints. Examine alternative causes, including incorrect answers or missing task information. Do not optimize a metric at the expense of correctness or control.

Signal boundary

List any proposed sensing separately from existing inputs. Specify opt-in, visibility, retention and who can access the data. Avoid employer-facing focus scoring.

End with one concrete interaction improvement and a planned evaluation, clearly separating the useful non-neural design from any unshipped sensor proposal.

skills

talk-pari-patel-agent-awareness

tile.json