Source: https://internal-agents.com/agents/dropbox-deflaker # Dropbox — Deflaker Deflaker investigates and repairs flaky tests for Dropbox engineers using Nova. - Company: [Dropbox](https://internal-agents.com/organizations/dropbox) - Collection: Agents - Approach type: Agent - Deployment stage: Deployed - Autonomy: Unknown - Evidence strength: Detailed primary - Status: Internal - First reported year: 2026 - Work: Coding - Invocation: Event-driven - Entry reviewed: 2026-09-17 ## Purpose ### Summary Deflaker investigates and repairs flaky tests for Dropbox engineers using Nova. Fact · Reported · High confidence · `dropbox-deflaker--summary` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned Representative workflow: Athena flaky-test detection → CI-validated fix or capped attempts. ## How it works ### Collect evidence Athena identifies flakiness; Deflaker gathers passing and failing logs. Fact · Reported · High confidence · `dropbox-deflaker--primitives-0` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned ### Propose a repair Nova diagnoses the likely cause and proposes a change. Fact · Reported · High confidence · `dropbox-deflaker--primitives-1` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned ### Retry or finish Failed validation starts another attempt with new logs and prior notes; the workflow lands a fix or stops after five attempts. Fact · Reported · High confidence · `dropbox-deflaker--primitives-3` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned ## Where people stay involved - **Athena flaky-test detection → CI-validated fix or capped attempts** — Unknown · Level unknown **Unreported:** The source describes an automated fix-and-validate loop and publication outside the agent, but does not state the required human approval or exception-handling policy. ### Supervision evidence #### Operating model assessment Unclassified for Athena flaky-test detection → CI-validated fix or capped attempts; human attention boundary: unknown. Inference · Catalog judgment · Unverified confidence · `dropbox-deflaker--operating-models-0` Confidence reason: The source describes an automated fix-and-validate loop and publication outside the agent, but does not state the required human approval or exception-handling policy. Qualifications: - Observation date: 2026 Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · Flaky test remediation: workflow lands a working fix or reaches five attempts; diagram describes the automatic retry loop ## Implementation details ### Harness Durable remediation workflow calling Nova; publication and CI remain outside the coding agent. Fact · Reported · High confidence · `dropbox-deflaker--architecture-harness` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned ### Context management Prior attempt notes and new test logs accompany retries. Fact · Reported · High confidence · `dropbox-deflaker--architecture-context-mgmt` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · How we’re using the platform: Flaky test remediation; What we learned ### Implementation coverage - **Model:** unreported — Not documented for this subject in the reviewed source. - **Harness:** reported - **Sandbox:** unreported — Not documented for this subject in the reviewed source. - **Tool access:** unreported — Not documented for this subject in the reviewed source. - **Knowledge:** unreported — Not documented for this subject in the reviewed source. - **Context management:** reported - **Credentials:** unreported — Not documented for this subject in the reviewed source. - **Interfaces:** unreported — Not documented for this subject in the reviewed source. ## Validation and failure handling ### Validate repeatedly CI repeats the test at least 100 times, depending on its failure rate. Fact · Reported · High confidence · `dropbox-deflaker--primitives-2` Confidence reason: A linked first-party source states the claim. Evidence: - Supports · [1] [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) · Flaky test remediation: CI runs the proposed change 100 or more times depending on failure rate ## Reported observations **Unreported:** The reviewed source does not document this for the named subject. ## Lessons **Unreported:** The reviewed source does not document this for the named subject. ## Question coverage and scope - **purpose:** Reported - **workflow:** Reported - **human involvement:** Unreported — The source describes an automated fix-and-validate loop and publication outside the agent, but does not state the required human approval or exception-handling policy. - **implementation:** Reported - **validation:** Reported - **observations:** Unreported — The reviewed source does not document this for the named subject. - **lessons:** Unreported — The reviewed source does not document this for the named subject. ## Related reading - Built on: [Dropbox — Nova](https://internal-agents.com/agents/dropbox-nova) ## Sources 1. [Introducing Nova, our internal platform for coding agents](https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents) - Engineering blog · First party · Evidence - Original URL: - Accessed: 2026-08-12 · Last verified: 2026-08-31