Describe the job.
Choose the depth.
A local matching board for the moment when a tiny coding robot needs more than code: a tool, a model, or a deeper pass of reasoning.
What are you trying to do?
Describe the outcome in your own words.
Best local fit
A transparent recommendation, not live discovery.
Your task includes research language, so a source-finding capability ranks highest.
A good tool feels like a continuation of the brief.
Find
Match the task language to the capability required, rather than browsing a store at random.
Think
Use a deeper reasoning mode when the task involves comparisons, uncertainty, or many steps.
Inspect
Leave a recommendation that explains its fit and its local-data boundary.
What the local capability catalog actually matches
Read the explanation
The source ranks four fixed local capability descriptions: research connector, browser automation, code workspace and planning board. It lowercases your task and counts how many of each category’s four keyword strings occur anywhere in it. Repeating the same keyword does not add another point, while a substring match can count inside a longer word. This does not discover installed plugins, check permissions or contact a retailer, browser agent or model. The default task contains research, Home Depot and compare. Those are three distinct research keywords, producing a score of three on this source’s local catalog. A task containing only research gives one, even if that word is repeated. These bars use fifty pixels per matched keyword: one hundred fifty and fifty pixels. When scores tie the source retains catalog order, and no match starts with the first research category as a fallback. Scores do not prove a capability exists or fits an account. Standard and deep are saved mode labels; changing them does not change the keyword algorithm or launch deeper reasoning. Task changes and mode changes save task and mode locally, and reload restores them. Export includes the full task, mode, top category, match score, every ranked category score and explicit source boundary. Reset returns to the original task and standard mode, then saves those defaults again. No plugin installation or backend model request is performed.