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

outline.mdskills/talk-fahmy-harness-api-design/references/

Source outline

Use transcript.md for exact wording. Promotion cues 156–166 are omitted.

TimeCuesTopic
00:00–01:531–32Background and evolution toward orchestration
01:58–04:1833–74Reusing harnesses and model/harness coupling
04:20–07:5075–130Multi-harness workflows and glue-code example
07:54–09:20131–155ACP, steering and workspace gaps
10:05–12:55167–210Release churn, interaction and workspace state
13:01–14:44211–242Extensibility, approvals, processes and errors
14:48–17:23243–283Bento, remote abstraction and ACP attempts
17:25–19:44284–320Minimal harness/workspace split and streaming
19:47–21:26321–346Loops and orchestration examples
21:28–24:06347–384AFK demo and unfinished portability
24:17–29:50385–473Deterministic orchestration versus agent choice
30:10–31:13476–493Message-history transfer and context limitations

Short quotations

  • “agents don't work in vacuum.” — excerpt, cue 202, 12:27.
  • “It only works for the use cases I built it for.” — cue 375, 23:37.
  • “every single harness handles its own context.” — excerpt, cue 481, 30:29.

Limits

The SDK is an evolving implementation/PR account. Handoffs preserve exported messages but not universal reasoning state or compaction behavior. ACP support and release counts are event-time observations. Model-family review benefits and editing-tool coupling are not quantified benchmark findings here. The 8,000-to-400-line comparison lacks a reproducible methodology in the transcript.

skills

talk-fahmy-harness-api-design

tile.json