autonoetic

Front-door lead agent for ambiguous goals.

Durable software engineering agent for reusable code and artifacts.

Lightweight execution agent for basic bash and dependency-free scripts.

Installs new durable agents into the runtime.

Cron-driven root orchestrator of the evolution pipeline: analyses sessions, triggers curator + steward, surfaces admin proposals.

Operator-triggered: decides whether a tactic proven in a session should become reusable, and by which route — instruction, wrapper, or new skill.

Generates wrapper agents for I/O gaps

Cross-session learning distillation and per-agent performance scoring for the evolution pipeline.

Research-focused autonomous agent for evidence collection.

Per-agent evolve/skip decision maker — triggers factory pipeline for agents needing improvement.

Builds and installs new agents end-to-end: design, code, package, gate, install.

Build-time dependency resolution and artifact layering agent.

Institutional office: works the anomaly-flag queue, chases adjudication SLA breaches, and adjudicates flags via the native anomaly_adjudicate tool.

Collaborative lead agent that uses PlanFrames for human-agent co-construction.

Builder agent that installs durable specialist workers from chat requests.

Builder agent for probing whether generated workers infer tiered memory architecture from intent.

Design, structure, and task decomposition agent.

Audit, review, and promotion gate agent.

Static code review for artifact correctness, credential flow, and behavioral contracts.

Runs artifact test suites in a no-network sandbox.

"Validation and testing autonomous agent (operator-invokable sealed-network evaluation)."

Debugging and root cause analysis agent.

A/B replay orchestrator: runs a set of tasks against two agent revisions and returns a statistical comparison report.

Independent post-session judge that grades a session's Completion based on its SessionOverview. Runs after the post-session digest and attaches a verdict to the session_outcomes row.

"Handles multi-step human-in-the-loop credential ceremonies (OAuth, identity verification, manual token entry) after the planner starts onboarding."

Observer agent that reviews agent sessions for trajectory divergence patterns.

Tool-free single-completion divergence judge used by the sentinel validation experiment.

"Handles multi-step human-in-the-loop credential ceremonies (OAuth, identity verification, manual token entry) after the planner starts onboarding."

Analyzes failed sessions with code-level root causes and files scoped GitHub issues.

Translates natural-language governance intent into structured constitutional amendment proposals.

Finds installed agents that match a given task intent using agent_list + LLM reasoning.

System-tier red-team agent: proposes adversarial attack patterns the sentinel should catch. Structurally separated from the sentinel — never authors eval suites targeting itself.

Frozen baseline sentinel — runs deterministic-only checks. The gateway-side SentinelRunner executes the checks without LLM calls; the llm_config here is a thin orchestrator fallback for structured output formatting only.

System-tier security sentinel: periodic read-only audit of agents, artifacts, and gateway state for security issues.

Builder agent that installs durable specialist workers from chat requests.

Registers to a remote service via credential.setup(skill_url) and human input.

Memory-first quickstart agent for gateway terminal chat.