Agents overview
bettersense includes 22 subagents alongside the 30 skills. They’re different tools for different situations.
Skills vs. agents
Section titled “Skills vs. agents”| Skills | Agents | |
|---|---|---|
| Context | Runs inside your current session | Gets a fresh context window |
| State | Carries your session history | Starts clean |
| Best for | Framework-guided conversations, quick structured outputs | Long documents, multi-pass analysis, tasks that compete with session state |
| Invocation | Auto-routes or /bettersense:<skill-name> | /agents then select, or Claude delegates automatically |
When to use an agent
Section titled “When to use an agent”Agents earn their keep when:
-
The task involves a large document. An RFC review, a complex spec, a 30-page strategy doc. Reading the whole thing in a session full of other context produces worse analysis than a fresh window.
-
The task is multi-pass. The eval designer runs several passes (problem statement → metric design → test cases → rubric → edge cases). Each pass benefits from the previous without competing with your meeting prep from earlier.
-
You want a second opinion. Fresh context means the agent hasn’t been anchored by your framing from earlier in the session.
How agents are invoked
Section titled “How agents are invoked”From within Claude Code:
/agentsSelect the agent from the list that appears. You can also describe what you’re working on and Claude may route to an agent automatically.
Explicitly by name:
use the-spec-writer for thisthe-eval-designer should look at this eval setupThe clusters
Section titled “The clusters”Seven agents covering the full lifecycle of shipping an AI feature — from problem validation through production reliability.
the-reducer → the-discovery-facilitator → the-scientist → the-eval-designer → the-architect → the-prompt-critic → the-red-teamer
Three agents for turning technical work into written artifacts — specs, explanations, and executive communication.
the-spec-writer · the-explainer · the-translator
Four agents covering the reliability arc — design review, reliability planning, incident response, and postmortems.
the-rfc-reviewer · the-slo-designer · the-incident-responder · the-postmortem-facilitator
Two agents for qualitative data synthesis and multi-team program management.
the-research-synthesizer · the-program-manager
Six agents covering business operations, team effectiveness, and career development.
the-vendor-evaluator · the-gtm-planner · the-retro-facilitator · the-data-storyteller · the-status-crafter · the-career-coach