Agent
Haiku web-research scout — runs scope.md's queries, vets accessibility, writes the shared source corpus. Discovery only.
5 2d ago A 29 tokens
Agent
Sonnet web-research specialist — deep-reads a scout's source corpus, verifies claims, writes claims.json + summary.md.
5 2d ago C 30 tokens
Agent
Opus NotebookLM sweep — blocked until workers finish, assesses claim coverage, fills gaps, frames the final document.
5 2d ago C 27 tokens
Agent
Opus web-research sweep — blocked until specialists finish, runs an adversarial coverage check, fills gaps, writes the summary.
5 2d ago C 32 tokens
Agent
Sonnet NotebookLM MCP worker — blocked until scout supplies sources; ingests them, queries, writes {letter}-claims.json. NotebookLM tool surface distinguishes it from research-scout/specialist/synthesizer and other -worker agents.
5 2d ago C 53 tokens
Agent
Applies a reviewer's findings to the target artifact with reasoning annotations; escalates disagreements instead of skipping them.
5 2d ago C 26 tokens
Agent
Sonnet static-security scanner — path traversal, injection, secret leakage in a diff. Source code, not dependency manifests.
5 2d ago A 29 tokens
Agent
Personas are Opus-only. The Front-End Reviewer reviews front-end code for design-system adherence — tokens, components, CSS architecture.
5 2d ago C 32 tokens
Agent
Personas are Opus-only. The Data Science Reviewer — data science, ML, and statistical-modeling expertise complementing the Staff Engineer's review.
5 2d ago C 35 tokens
staff-eng
34Agent
Personas are Opus-only. The Staff Engineer — uncompromising staff-engineer review of code, plans, architecture, docs. The generalist reviewer.
5 2d ago A 35 tokens
staff-ux
35Agent
Personas are Opus-only. The UX Reviewer reviews user-facing flows for clarity, trust signals, and intuitive design.
5 2d ago C 28 tokens
Agent
Opus structured-research synthesizer — reconciles verifier findings into schema-conforming YAML/JSON, never prose.
5 2d ago C 29 tokens
Agent
Classifies EM-captured test failures (real/flake/env/timeout/skip) to disk. Mechanical triage only, no opinions.
5 2d ago A 33 tokens
test-runner
38Agent
Runs a diff's scoped tests, reports raw evidence. Tier-T only, never the fast/full suite.
5 2d ago A 25 tokens
vp-product
39Agent
Personas are Opus-only. The VP-Product Reviewer (they/them), VP Product — stress-tests choices before they ship: refactor-over-patch, alternative shapes.
5 2d ago C 39 tokens
Command
Build or refresh the architecture atlas via scout, analyst, synth.
5 2d ago C 17 tokens
autonomous
41Command
Toggle autonomous mode — suppresses handoff nudges near compaction.
5 2d ago A 17 tokens
bug-blitz
42Command
Grind the bug backlog and tests; fix small, surface big items to PM.
5 2d ago A 21 tokens
cruft-sweep
43Command
Scan for reclaimable scratch and orphans; apply only if confirmed.
5 2d ago A 19 tokens
distill
44Command
Distill session artifacts to wiki and decisions; archive specs, drop scratch.
5 2d ago A 18 tokens
install
45Command
Installs the coordinator plugin — checks prereqs, configures project.
5 2d ago C 14 tokens
Command
Autonomous backlog run — flight-recorder prep, then run-through.
5 2d ago A 18 tokens
Command
PM-GATED, never from a subagent. NotebookLM research for video/audio sources.
5 2d ago B 18 tokens
research
48Command
PM-GATED, never from a subagent. Deep research — web, repo, or structured.
5 2d ago B 20 tokens