10 results
Jev-judged context compaction for DeepSeek Harness. Two layers: prune stale tool results (semantic instead of size-only), and replace spent tool-call ranges with a deterministic receipt instead of a model-written summary.
Skill-trigger evaluation: an LLM judge recreates the DSH skill catalog and measures how reliably a skill description routes matching queries.
Commitment write-gate for AI coding agents: two-tier (deterministic + LLM judge) pre-execution policy. Engine-agnostic core with a DeepSeek Harness (dsh) adapter.
Codex-style approval autopilot for DeepSeek Harness: ordered glob rules (allow/ask/deny) plus an AI risk judge (low/medium/high) mapped through a risk tolerance, as an approval answerer.
误食催吐判断
Cross-border counterparty due diligence for DeepSeek Harness — resolve the entity you are actually contracting with, then judge whether the record is fit to rely on.
Run one command N rounds and judge by median/distribution instead of a single run
Adversarial pro/con debate with a neutral judge for DeepSeek Harness: /debate human command, subagent-driven rounds, verdict record, and a web debate card.
Jev quality bench for DeepSeek Harness: shadow-accounted dangerous-command judging, injection screening for fetched content, canary A/B drills, and an optional gate mode — with the ledger, the report and the settings card needed to judge whether any of it is worth keeping.
A judgment kernel for coding agents: bounded questions a small Jev judge answers, so the main model keeps its context and tokens for real reasoning.