Build Agent Evals
Use when an agent needs the smallest useful eval pack to catch regressions or false confidence early.
Evolve Skill Library
Use when the skill library feels crowded, redundant, stale, or mis-scoped and needs pruning or reshaping.
Set Agent Autonomy Boundaries
Use proactively when an agent lacks checkpoint policy or risky actions need clear human-review boundaries.
Trim Agent Scope
Use proactively when a plan, skill, or agent keeps absorbing side quests and needs to return to core scope.
Choose Agent Architecture
Use when an agent's job is clear and you need to choose structure, boundaries, or single-agent vs multi-agent.
Design Agent Memory
Use when deciding what an agent should remember across turns or sessions versus fetch on demand.
Design Agent Tools
Use when a tool is misused or underspecified because its inputs, errors, or side effects are unclear.
Verify Agent Readiness
Use proactively before declaring success, shipping, or moving on when an agent or change needs fresh verification evidence.
Design Agent Permissions
Use after autonomy policy is set and you need permission classes, defaults, approvals, and denial behavior.
Design Agent Context
Use when an agent reads too much, misses key files, or needs a clear eager-vs-deferred context plan.
Orchestrate Agent Subagents
Use when delegation is already justified and you need safe subagent ownership, context, and merge boundaries.
Turn Postmortem To Improvement
Use after a failure, correction, or standout success to decide what reusable guardrail, eval, skill, or rule should follow.
Route Agent Design
Use proactively when a user is defining, constraining, debugging, or evaluating an agent system and needs the right PromethOS skill.
Define Agent Requirements
Use when an agent idea is still fuzzy and v1 scope, success bar, or operating boundaries are not concrete.
Plan Agent Implementation
Use when architecture is settled and you need an ordered implementation plan with checkpoints and verification.
Debug Agent Failures
Use proactively when an agent repeats a failure, loops, or degrades and needs a local root-cause check before redesign.