菜单
Orchestrator for Truesight MCP skills. Use this when the user needs help choosing the right Truesight workflow or when intent is ambiguous across LLM evaluate, error analysis, review, templates, or evaluation creation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Scope what quality should be measured, convert it into one or more actionable binary evaluations, deploy those evaluations through Truesight MCP, and generate a companion skill that applies them correctly. Use when a user wants to create new evals, quality checks, guardrails, or pass/fail criteria for AI outputs.
Generate synthetic test data for LLM evaluations using dimension-based tuple expansion. Use when the user needs synthetic traces, test cases, eval datasets, or when create-evaluation needs synthetic fallback data.
Fastest route to a deployed live evaluation using a pre-built Truesight template. Use when the user wants a quick start without building judgment configs from scratch.
Build a custom web interface for trace annotation and review. Use when users need a bespoke review surface for their workflow.
Systematically identify and categorize failure modes in evaluated traces using Truesight datasets and error-analysis tools. Use when quality issues are unclear, after major pipeline changes, or when incidents indicate drift.
Audit an existing evaluation workflow and produce severity-ranked findings with concrete next actions. Use when inheriting an eval setup, diagnosing quality regressions, or checking LLM evaluation process maturity.
| name | truesight-workflows |
| description | Orchestrator for Truesight MCP skills. Use this when the user needs help choosing the right Truesight workflow or when intent is ambiguous across LLM evaluate, error analysis, review, templates, or evaluation creation. |
Use this skill as the routing entrypoint across all Truesight MCP skills.
This skill is a router. It decides intent and routes to exactly one skill path.
Do not execute deep workflow steps here unless the user already asked for a very specific action and no further routing is needed.
create-evaluationevaluate-traceerror-analysisreview-and-promote-tracesbootstrap-template-evaluationeval-auditbuild-review-interfacegenerate-synthetic-dataWhen user intent is unclear, ask one question at a time using the structured question tool (loaded per the HARD-GATE above). Structure each with a short header, options with labels and descriptions, and place the recommended option first. Do not add "(Recommended)" or similar annotations to option labels.
Question format:
Which workflow do you want to run first?
A) Evaluate traces with an existing live eval
B) Run error analysis on a dataset
C) Review and promote flagged traces
D) Bootstrap from a template
E) Create a new evaluation from scratch
F) Audit my eval setup
G) Build a custom review interface
H) Generate synthetic test data
Rules:
create-evaluation, do not decompose it into smaller skills.