internal/

Package agent orchestrates model turns, tool approvals, and tool results.

Package collaboration exposes a bounded set of sub-agent collaboration tools.

Package credential resolves model API credentials without coupling command execution or environment precedence to the CLI bootstrap.

Package hook runs trusted lifecycle commands without invoking a shell.

Package instructions discovers directory-scoped project instruction files.

Package mcp connects trusted Model Context Protocol servers and exposes their tools through SuperCode's provider-neutral registry.

Package memory manages file-backed, cross-session long-term memory.

Package modelcatalog keeps model capabilities out of provider adapters and lets OpenAI-compatible endpoints describe their own model IDs in YAML.

Package permission models explicit, scoped access grants independently from the approval UI and the operating-system sandbox implementation.

Package plugin loads trusted local extension bundles.

Package policy stores explicit, user-approved execution rules.

Package provider defines the model-provider boundary used by SuperCode.

Package session stores resumable SuperCode conversations as local JSON.

Package skill discovers SKILL.md instructions and injects explicitly named skills into a model turn.

Package taskstate provides model-managed plans and explicit long-term goals.

Package tool defines SuperCode's model-callable tool boundary and registry.

Package userinput provides a blocking model tool backed by an interactive UI.