Automated README summary
Automatically summarized from the README on 27 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A DeepSeek Harness plugin that classifies each non-read-only tool call's risk, irreversibility, task-fit and injection suspicion, and re-scores recalled session messages by value instead of dropping oldest-first. Aimed at agent users who want fast, cheap, fail-closed decisions.
How it uses Jev
Before a tool call runs, Jev answers a Choice risk class plus Noul questions on irreversibility, task-match and injection, fed bounded argument previews in the state data field. Confident high-risk calls are denied, ambiguous ones escalated to a human, clearly-granted reversible ones auto-approved behind argument-evidence gating; a per-message Score re-ranks session-reference retention. Doubts degrade to stock behavior.
Primitives:choicenoulscore
Technique worth stealing
Use a non-generative typed classifier for per-call risk decisions, with shadow mode and fail-closed fallback to stock behavior before enforcing.
Try it
Install with `dsh plugin --profile add dsh-jev-interceptor`, start in shadow mode, inspect `/jev-stats`, then `/jev-mode enforce`.
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.14.0- Project signals
jev-1.14.0- Category
jev-1.14.0- Pattern tags
jev-1.14.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.94
- Shows a System One patternFound in material0.97
- Describes handling uncertaintyFound in material0.91
- Reports measured numbersnot found in the excerpt0.30
- Is a runnable toolFound in material0.98
- Top-10 learning shortlist signalFound in material0.63
- Is a replica of the modelnot found in the excerpt0.02
- Solves a common problem (0–2)score on a 0–2 scale1.28 / 2
- About JevFound in material0.99
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 27 Oct 2026.