TIKAZ中文
← Back to the Skill catalog

⚡ Independently installable TIKAZ Skill

context-orchestrator

Route long text, files, webpages, code, tables, images, audio, video, or long conversations between default Context Economy and explicit ON-triggered Context On-Premise while keeping one evidence and cost contract.

Workflow group: Context EconomySpecialist Skill

Evidence and core advantages

2 routesone deterministic controllerEconomy by default; On-Premise only for a valid ON prefix
5 formsfalse-positive trigger guardon, ONLY, ON-board, mid-sentence ON, and Chinese-prefixed ON stay Economy
1 contractquality and cost stay separateevidence slots, omissions, Pending, fallback, and ledgers use one schema

Fixed public benchmark; tokens are deterministic estimates, not provider billing. Short inputs may grow from protocol overhead. Results are not universal accuracy or cross-project superiority claims. Open the full evidence card →

01

What it owns

Route long text, files, webpages, code, tables, images, audio, video, or long conversations between default Context Economy and explicit ON-triggered Context On-Premise while keeping one evidence and cost contract.

02

When to use it

Install and invoke this Skill directly when its named capability is the complete task. Use the suite orchestrator when the outcome crosses multiple capabilities.

03

Install independently

Copy only this Skill folder. No sibling Skill is required. Keep the destination folder identical to the Skill name.
Copy-Item -Recurse -LiteralPath '.\suites\context-economy\context-orchestrator' -Destination '.\.agents\skills\context-orchestrator'

04

Try it directly

ON:把这个扫描 PDF 和视频整理成可追溯的 Context Pack,并分别报告命中、保真和估算 Token。

05

Execution contract

The single English execution source owns inputs, outputs, validation, fallback, and limits. User documentation never creates a second executable rule set.Open the execution source ↗

Skills in this group

context-orchestratorcontext-economycontext-onpremisecontext-packconversation-checkpointcontext-auditcontext-benchmark