Per the "How to Contribute" guidance (create an issue for missing papers), suggesting an addition.
Section: 3.3 Code Agents (next number, 123)
Entry:
123. Harness Engineering: "Harness Engineering for Language Agents: The Harness Layer as Control, Agency, and Runtime" [2026-03] [paper]
Why it fits: It formalizes the agent harness layer — the runtime/control loop wrapping tool execution that powers coding agents (SWE-agent, OpenHands, etc.) — as Control, Agency, and Runtime (CAR), audits 63 harness-relevant works, and proposes HarnessCard for reporting. Relevant to how code agents are built and evaluated.
Disclosure: I am an author. Happy to open a PR instead if you prefer. (DOI: 10.20944/preprints202603.1756)
Per the "How to Contribute" guidance (create an issue for missing papers), suggesting an addition.
Section: 3.3 Code Agents (next number, 123)
Entry:
123. Harness Engineering: "Harness Engineering for Language Agents: The Harness Layer as Control, Agency, and Runtime" [2026-03] [paper]
Why it fits: It formalizes the agent harness layer — the runtime/control loop wrapping tool execution that powers coding agents (SWE-agent, OpenHands, etc.) — as Control, Agency, and Runtime (CAR), audits 63 harness-relevant works, and proposes HarnessCard for reporting. Relevant to how code agents are built and evaluated.
Disclosure: I am an author. Happy to open a PR instead if you prefer. (DOI: 10.20944/preprints202603.1756)