https://speakerdeck.com/raeminkang 2026-11-03 00:22:49 -0500 Slides from DroidKnights 2026 (Nov 2, 2026) by Raemin Kang & Dayeon Baek, Yeogi Eottae. (Slides in Korean) How our Android team made AI coding agents work the team's way — from start to finish, with proof. 1. Harness — a frame that keeps the AI on the team's workflow until the job is done - v1: shared team skill, Stop hook, test-gated Phase/Step - v2: per-repo config picks the checks; scripts, not the AI, decide pass/fail 2. LLM Wiki — recording team decisions that never make it into code, so the AI reads them before it works 3. Verification — not trusting "Done!" from the AI - Convention: andcon (Konsist, detekt, ktlint) - UI: Roborazzi + Figma value matching - Behavior: Maestro E2E — DroidKnights 2026 발표 자료입니다. AI 코딩 에이전트를 안드로이드 개발에 도입하며 하네스, LLM Wiki, 검증 도구로 문제를 푼 과정을 공유합니다. Slides from DroidKnights 2026 (Nov 2, 2026) by Raemin Kang & Dayeon Baek, Yeogi Eottae. (Slides in Korean) How our Android team made AI coding agents work the team's way — from start to finish, with proof. 1. Harness — a frame that keeps the AI on the team's workflow until the job is done - v1: shared team skill, Stop hook, test-gated Phase/Step - v2: per-repo config picks the checks; scripts, not the AI, decide pass/fail 2. LLM Wiki — recording team decisions that never make it into code, so the AI reads them before it works 3. Verification — not trusting "Done!" from the AI - Convention: andcon (Konsist, detekt, ktlint) - UI: Roborazzi + Figma value matching - Behavior: Maestro E2E — DroidKnights 2026 발표 자료입니다. AI 코딩 에이전트를 안드로이드 개발에 도입하며 하네스, LLM Wiki, 검증 도구로 문제를 푼 과정을 공유합니다. Tue, 03 Nov 2026 00:00:00 -0500 https://speakerdeck.com/raeminkang/droidknights-workflow-ax-d1979c40-85ff-4b3a-a09d-36a45bee1e89 https://speakerdeck.com/raeminkang/droidknights-workflow-ax-d1979c40-85ff-4b3a-a09d-36a45bee1e89 Slides from DroidKnights 2026 (Nov 2, 2026) by Raemin Kang & Dayeon Baek, Yeogi Eottae. (Slides in Korean) How our Android team made AI coding agents work the team's way — from start to finish, with proof. 1. Harness — a frame that keeps the AI on the team's workflow until the job is done - v1: shared team skill, Stop hook, test-gated Phase/Step - v2: per-repo config picks the checks; scripts, not the AI, decide pass/fail 2. LLM Wiki — recording team decisions that never make it into code, so the AI reads them before it works 3. Verification — not trusting "Done!" from the AI - Convention: andcon (Konsist, detekt, ktlint) - UI: Roborazzi + Figma value matching - Behavior: Maestro E2E — DroidKnights 2026 발표 자료입니다. AI 코딩 에이전트를 안드로이드 개발에 도입하며 하네스, LLM Wiki, 검증 도구로 문제를 푼 과정을 공유합니다. Slides from DroidKnights 2026 (Nov 2, 2026) by Raemin Kang & Dayeon Baek, Yeogi Eottae. (Slides in Korean) How our Android team made AI coding agents work the team's way — from start to finish, with proof. 1. Harness — a frame that keeps the AI on the team's workflow until the job is done - v1: shared team skill, Stop hook, test-gated Phase/Step - v2: per-repo config picks the checks; scripts, not the AI, decide pass/fail 2. LLM Wiki — recording team decisions that never make it into code, so the AI reads them before it works 3. Verification — not trusting "Done!" from the AI - Convention: andcon (Konsist, detekt, ktlint) - UI: Roborazzi + Figma value matching - Behavior: Maestro E2E — DroidKnights 2026 발표 자료입니다. AI 코딩 에이전트를 안드로이드 개발에 도입하며 하네스, LLM Wiki, 검증 도구로 문제를 푼 과정을 공유합니다. Tue, 03 Nov 2026 00:00:00 -0500 https://speakerdeck.com/raeminkang/droidknights-workflow-ax https://speakerdeck.com/raeminkang/droidknights-workflow-ax