Positron Settings

Positron settings

These commands read the running Positron's configuration -- what the user has set, what this build's settings registry knows, and which AI model providers are set up. Two rules frame everything in this skill:

1. This is state, not documentation. The user's configuration is live state in this running Positron. Do not fetch a documentation page, do not read release notes, and do not read a settings.json file to answer what is configured or what is in preview: none of those can know what this install has. Query it.
2. Never claim you cannot inspect settings. You can, with the commands in this skill. When the payload has a limit, say what it does and does not show (the reference file spells those limits out) instead of deflecting to "I can't see your settings from here" or sending the user to click through the UI for an answer you can query.

Calling these commands

Invoke commands with the positronCommand tool, passing the command's literal id exactly as written in the reference files -- copy it, do not retype it from memory. Where a command takes arguments, fill args positionally in the order given under that command's "Arguments" entry. Omit args entirely when using none -- do not pass an empty object or array. Never invent an argument value the user hasn't given you or that isn't documented.

When a command doesn't work

• not-found: the command id isn't present in this Positron build, meaning the build is older or newer than this skill expects. Report this plainly; do not substitute a similarly named id you're unsure about.
• No command here has a precondition: there is always an answer, even when the user has set nothing and no provider is configured, so a disabled failure is unexpected. Report it plainly if it ever appears.

When a result is truncated

Truncation is a field to read, not a shape to infer. If a command response has a top-level truncated object, the transport cut the result to fit its budget and is telling you so: truncated.field names the list that was cut, and you received truncated.returned of truncated.total entries. Say how many entries are missing rather than presenting the list as complete, and narrow the request (a filter, a limit, explicit keys) to get the rest. If truncated.preview is true, result is cut JSON handed to you as a string: do not parse it, say the result was too large, and narrow the request. No truncated object means nothing was cut.

Reference Files

Configuration -- references/configuration.md Read when the user asks which settings they have configured, what a setting is set to, why a setting they set is not taking effect, which features are in preview or experimental, whether a setting exists, or what a setting's default or allowed values are. Also carries the known settings-that-gate-other-settings relationships (the AI feature switches), which no single payload can reveal.

Providers -- references/providers.md Read when the user asks which AI model providers they have set up, enabled, or signed in to, whether a provider is usable, why a provider stopped working, or where provider configuration lives.