Agents
The Agent Roster
Last updated July 26, 2026
Agents
The kit ships a roster of global agents. The pipeline spawns them by name via agent(...). Each agent's model comes from its own frontmatter chain, except where the pipeline overrides it with a call-site model= (the reviewers and the ultra seats).
| Agent | Model | Purpose |
|---|---|---|
planner |
genius tier (@plan) |
Architect. Three modes: CLARIFY (question tree), PLAN (structured plan), CONSULT (adjudicate a stuck builder's design question). Investigates only via cheap subagents; never implements. |
task |
@task (the task pool) |
Mechanical worker. Does the actual implementation, offloads research to scouts, and returns a stuck signal instead of thrashing. |
deep-debugger |
genius tier (@slow) |
Root-cause diagnostician. Read-only; returns the cause plus the exact fix and how to verify. Does not implement. |
deep-reviewer |
genius tier, varied per call | Clean reviewer with no native output schema, so call-site schemas apply cleanly. The pipeline overrides its model per lens (reviewers diversity set) and per ultra seat. |
designer |
@designer |
UI/UX specialist. Builds, modifies, and improves frontend pieces, reviews the design lens, and fixes frontend findings. Design-system-first and accessibility-aware. |
david-research |
@smol |
Cheap external research scout (web, docs, repos, APIs). Keeps internet context out of the parent. |
review-orchestrator |
genius tier (@slow) |
Legacy manual review-loop orchestrator. Superseded by the in-pipeline run_review_loop(); kept for standalone, non-eval use. Can drive a three-family panel via the pinned reviewers below. |
fable-reviewer |
Fable 5 (pinned, high) | Model-pinned panel reviewer for manual cross-family reviews. Read-only; not used by the pipeline's review loop. |
sol-reviewer |
GPT-5.6 Sol (pinned, xhigh) | Model-pinned panel reviewer. Read-only; not used by the pipeline's review loop. |
opus-reviewer |
Opus 5 (pinned, max) | Model-pinned panel reviewer. Read-only; not used by the pipeline's review loop. |
The genius-tier agents reference roles (@plan, @slow) in their frontmatter, so retuning the anchored genius list in modelRoles moves all of them at once. The reviewers and every read-only investigator can spawn david-research to offload external docs and API lookups instead of burning their own context on the internet. The exact model strings live in the modelRoles config and in each agent's frontmatter, and both are tunable. See Configuration.
Project agents
A repo can vendor its own specialists in .omp/agents/*.md. The pipeline discovers them at plan time, adds them to the plan schema's agent choices, and lists each one with its description in the plan prompt, so the planner assigns pieces to them when the work falls squarely in their documented domain. Each runs with its own frontmatter model, skills, and spawn whitelist. Kit personas are excluded from discovery, and a repo without .omp/agents gets the base choices unchanged. Project agents do not join the task pool rotation.
Bundled omp scouts
The planner and the genius agents also spawn omp's own bundled scouts, which are not part of this kit: scout (read-only local codebase scout; named explore before omp 17) and librarian (library and API source). Both run on the cheap smol role. These keep fact-finding off the expensive genius reasoning.
Agent selection is an invariant, not a preference. Plan and consult go to planner, hard diagnosis to deep-debugger, review to deep-reviewer, always via an explicit agent=. Persona text in the prompt never substitutes for agent= (omp ≥17 removed the task tool's old role= field outright). See Resume and recovery for why this matters.