Graphorin API reference v0.7.0
Graphorin API reference / @graphorin/skills / registry
registry
SkillRegistry - registry over loaded skills.
The registry is the surface the agent runtime (Phase 12) and the standalone server (Phase 14) consume. It exposes:
getMetadata()- every skill's Tier-1 metadata, used by the ContextEngine to assemble the system prompt's skill metadata block (Phase 10d).activate(triggers)/getActivationRequest(triggers)- match a list of trigger strings (slash commands and / or model- emitted skill names) and return the corresponding ActivatedSkill records.getSkill(name)- direct lookup.tools()- flat list of declared tool entries; the runtime resolves the actualTool[]through the@graphorin/toolsregistry.
Interfaces
| Interface | Description |
|---|---|
| ActivationRequest | Activation request produced by SkillRegistry.resolveTrigger. |
| ParsedActivationTrigger | Parsed activation trigger. The registry uses this to discriminate slash-command activations (which override disable-model-invocation: true) from model-emitted auto activations (which honour it). |
| RegisteredToolDeclaration | Tool-declaration record exposed by SkillRegistry.toolDeclarations. Adds the owning skill's name and trust level so downstream registrations into @graphorin/tools can stamp the source. |
| SkillRegistry | Public registry surface. |
| SkillRegistryOptions | Options accepted by createSkillRegistry. |
Type Aliases
| Type Alias | Description |
|---|---|
| SkillToolStamper | Stamping seam injected by the agent runtime (Phase 12). It turns a skill's pre-built Tool into a fully resolved ResolvedTool (trust class + sandbox tier + source). The skills package keeps no hard dependency on @graphorin/tools; when no stamper is configured, activate() surfaces no tools (the runtime resolves them itself). |
Functions
| Function | Description |
|---|---|
| createSkillRegistry | Build a fresh, empty registry. Multiple registries can co-exist within a single process; the framework defaults to a single shared instance per agent instance. |
| parseActivationTrigger | Parse a single activation trigger. Slash-command bodies (/skill:<name>) are routed through the slash parser; bare names are treated as auto-activation requests emitted by the model. |
References
StampedSkillTool
Re-exports StampedSkillTool
stampSkillTool
Re-exports stampSkillTool
stampSkillToolFromMetadata
Re-exports stampSkillToolFromMetadata