SkillKit.Agent (SkillKit v0.1.0)

Data struct and parser for AGENT.md files.

An agent carries all the configuration needed to start an agent process: identity, system prompt, skills, runtime, scope, and mailbox tuning.

The struct serves as the single source of truth flowing through the entire supervision tree.

Summary

Types

t()

Functions

parse(content)

Parses AGENT.md content into a %SkillKit.Agent{}.

Types

t()

@type t() :: %SkillKit.Agent{
  caller: pid() | nil,
  conversation_store: {module(), keyword()} | nil,
  depth: non_neg_integer(),
  description: String.t(),
  initial_messages: [term()],
  mailbox: %{max_messages: pos_integer(), flush_interval: pos_integer()},
  max_agent_depth: non_neg_integer(),
  model: String.t() | nil,
  name: String.t(),
  parent_ref: SkillKit.AgentRef.t() | nil,
  registry: atom() | nil,
  runtime: {module(), keyword()},
  scope: term(),
  skills: [{module(), keyword()}],
  system_prompt: String.t(),
  tools: [{module(), keyword()}]
}

Functions

parse(content)

@spec parse(String.t()) :: {:ok, t()} | {:error, term()}

Parses AGENT.md content into a %SkillKit.Agent{}.

Returns {:ok, agent} or {:error, reason}.