Agents

Reference catalog for Tahyi’s founding specialist agents. Each agent owns one operational domain and integrates with the coordination layer through a shared contract.

For the conceptual model, see Core Concepts — Specialist agents.

Founding specialists (1.0)

Agent ID	Domain	Owns	Example work
deployment-maintainer	Deployment	Release pipelines, rollouts, rollbacks	Promote a canary, roll back a bad deploy
monitoring-maintainer	Monitoring	Alerts, SLOs, on-call routing	Tune alert noise, acknowledge incidents
observability-maintainer	Observability	Logs, metrics, traces	Find root cause across signals
dba-maintainer	DBA	Databases, migrations, backups	Run a verified migration, restore a snapshot

Agent lifecycle

disabled → enabled (dry-run) → enabled (live) → idle / planning / executing / blocked

State	Meaning
disabled	Agent not registered for this environment
idle	Enabled, no active task
planning	Building an action plan from current context
executing	Running approved two-way actions
blocked	Waiting on approval gate, missing credentials, or external dependency

Enable an agent

npx tahyi agent enable deployment-maintainer \
  --environment staging \
  --dry-run

Remove --dry-run only after reviewing plans in the audit log and configuring production policy.

Safety scope per agent

Every agent operates under the same safety model:

Door type	Agent behavior
Two-way	Act autonomously within policy
One-way	Plan, log, wait for approval

Domain-specific examples:

Agent	Two-way examples	One-way examples
Deployment	Scale replicas, restart pods	Delete a deployment, purge release history
Monitoring	Silence a noisy alert	Delete an SLO definition
Observability	Run a read-only query	Delete log retention policy
DBA	Read-only schema inspection	DROP TABLE, destructive migration

Replaceability

Specialists are designed to be swappable:

• Each agent has a narrow tool scope and prompt boundary
• Handoffs use explicit task contracts — not shared mutable state
• Removing one agent should not break unrelated domains

If you need to fork or replace a specialist, keep the agent ID stable and swap the implementation behind it.

Handoff patterns

Common cross-agent flows:

Trigger	From	To	Task
Alert fires	Monitoring	Observability	Correlate logs/metrics for root cause
Root cause found	Observability	Deployment	Execute rollback
Migration planned	DBA	Deployment	Coordinate maintenance window rollout
Deploy completes	Deployment	Monitoring	Verify SLO health post-release

Handoffs create new tasks in the coordination layer — each with its own audit trail.

Out of scope (1.0)

These domains are not founding specialists:

• SecOps / security scanning
• FinOps / cost optimization
• Networking / DNS / CDN
• General-purpose coding agents

See SCOPE for the full boundary.

Related reading

• Core Concepts — swarm model and safety gates
• How it Works — execution loop and handoffs
• Architecture — coordination layer reference
• Configuration — agent and environment settings