Module agent_resource

Expand description

Standalone executable-agent declarations. Loading is explicit: existing MCP prompt registries and unrelated commands do not discover these resources.

Structs§

AgentResource
AgentSettings
DelegationSettings: Explicit agent targets that the runtime may resolve for agent.invoke. Empty settings are omitted from version snapshots for backwards compatibility.
ExecutionLimits
LoadedAgentResource
ModelDefinition: A declarative model alias. Credentials are referenced, never read here.
ModelInspection: Sanitized effective model settings for static inspection.
Permissions: Omission means read-only. An explicit allow list replaces that default. Capability enforcement belongs to the runtime, not prompt projection.
ProfileEntry: Combined profile discovery view. Profiles remain prompt-only and need no model.
PromptSettings
ResourceDirectory

catalog
cli_resource_directories
identifier 🔒
inspect_model 🔒
list: List executable standalone agents only. Prompt-only profiles are exposed by the separate ctx profile command.
list_profiles
load_resources: Read each layer in sorted order and replace whole resources only with an explicit override. A duplicate name within one layer is always an error.
resource_catalog 🔒
resource_directories: Choose resource paths using the same explicit-config isolation as settings. The caller supplies the workspace root, never inferring it from a DB path.
show: Show one executable standalone agent.
show_profile
test_profile
validate: Validate executable agent resources and their model references.
validate_builtin_models 🔒
validate_profiles