Use Jev to choose among available tools, skills, or agents, or assess outputs against explicit criteria. Receive structured selections, scores, or probabilities to support decisions within a defined set of options.
中文简介正在准备中。当前显示英语简介。
Use Jev to select among available tools, skills or agents, or evaluate outputs against explicit rubrics. Use when the user requests Jev or a bounded decision would help; not for open-ended writing or research.
安装前可以查看交给 AI 代理的指令内容。
Jev returns structured judgments. The current Agent retains planning, parameter construction, execution and native subagent collaboration.
jev-decision-model with ipollowork_extension_list_actions, then use ipollowork_extension_call with action evaluate and the payload below in args.jev_evaluate with the same payload. jev_status checks local configuration; jev_check_connection verifies the key with a small billable request.status and check-connection. A configured key is not proof of a successful API connection. If missing or invalid, direct the user to plugin authorization (Work) or the TYPESAFE_API_KEY credential (DSH); do not request the key in chat.Discover actual candidates from the host or the visible tool/skill catalog. Respect the user's explicit selections. Include only available, permitted candidates and their stable IDs, short descriptions and relevant constraints. Jev does not discover tools, install plugins, grant permissions or invoke another Agent by itself.
Send the task and minimal relevant context as state; do not send credentials, entire transcripts or complete files unnecessarily. Each question contains type, a complete instructions question, and the appropriate criteria. Question IDs identify results; they are not instructions to the model.
choice: choose one of 2–255 keyed candidates; include an other / none candidate when needed.score: grade against 2–10 ordered levels, indexed from zero. Several independent scores can rank candidates.noul: estimate the probability of a clearly stated yes/no condition. To select multiple complementary skills, batch one relevance question per candidate instead of forcing a single Choice. A value near 0.5 means uncertainty, not medium skill.Batch independent questions sharing a state; use a later request if a question depends on an earlier result. Plugin bounds are 64 questions and 256 KiB per request. These are plugin limits, not claims about the provider's limits.
{
"state": { "task": "整理已收集的竞品资料，制作汇报 PPT", "available": ["research", "slides"] },
"questions": {
"use_research": { "type": "noul", "instructions": "这项任务是否需要分析竞品资料？" },
"use_slides": { "type": "noul", "instructions": "这项任务是否需要制作演示文稿？" }
}
}
Preserve the returned probabilities, confidence, model and usage when reporting. Jev does not return written reasoning: distinguish your explanation from its structured result. Check current availability again before execution. Read the selected Skill and construct tool arguments using the current host's schemas.
Thresholds and weights come from the user's task or application policy, not universal magic numbers. Low confidence, missing options, or API failure calls for ordinary Agent reasoning or clarification; say when Jev was unavailable instead of inventing its result. A recommendation does not authorize external actions, override the user's choice or bypass approval. No automatic global routing or self-training is installed by this plugin.
Reference: https://docs.typesafe.ai/api
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Search the private research library in Acme Research and summarize relevant material. Include each returned document's title and source URL so readers can identify the sources and check the material behind the summary.
中文简介正在准备中。当前显示英语简介。
Capture product screenshots for landing pages, newsletters, social posts, and presentations by operating the real app. Check page structure, image pixels, and visual appearance, then retake images with overlapping text, clipping, or other visible defects.
中文简介正在准备中。当前显示英语简介。
Find the right official documentation before answering technical questions. It identifies the library and version, retrieves focused references and examples, and explains relevant APIs while separating documented behavior from inference.
中文简介正在准备中。当前显示英语简介。
Implement or upgrade code using the project's exact dependency versions and matching Context7 documentation. Check available APIs, make the smallest compatible change, run package checks, and identify behavior that still needs verification in a running application.
中文简介正在准备中。当前显示英语简介。
Create an OpenCode plugin for iPolloWork with custom tools and hooks, which run code at specific events. The skill builds the TypeScript file and adds it to the configuration so it can load after a session restart.
中文简介正在准备中。当前显示英语简介。
Launch a standalone Chrome browser in a Daytona sandbox and control it through CDP, a browser debugging connection. Use it to complete sign-in, OAuth authorization, and web administration flows, then verify the visible results.
中文简介正在准备中。当前显示英语简介。