Organize the lifecycle around a cast of distinct named expert personas, one per role, each with a persistent character — a name, an icon, named influences, and stated principles — that “shape decisions across all dispatched workflows.” Rather than invoking a neutral “planning command,” you talk to Mary the analyst or Winston the architect, and the persona colours how the work is done. This is bmad’s signature contribution: six named agents (Mary/Paige/John/Sally/Winston/Amelia) anchor its four phases. gstack is the second framework built the same way — its whole toolkit is a cast of role specialists (a CEO, an eng manager, a senior designer, a DX lead, a staff engineer, a QA lead, a chief security officer, an SRE, a release engineer) invoked as slash-commands.

The payoff beyond flavour is multi-perspective tension. BMAD’s Party Mode summons several personas into one roundtable where “the personas hold different priorities. The Architect guards the design, the PM guards scope, the Dev guards what’s actually buildable” — and can spawn “a separate agent for each persona every substantive round” so “no single mind colors them all.” That makes persona-agents a substrate for honest tradeoff decisions, pre-mortems, and reviews.

BMAD:

gstack (a role specialist per skill; the plan-review panel is the clearest cluster):

pstack — one instance, and the framework argues against the mechanism everywhere else:

  • pstack-comment-sicko — a read-only comment reviewer written in a deliberately deranged first-person voice (“I hate comments… everything else is meat”), the wiki’s most extreme persona and the furthest from the professional-role register gstack and bmad use.

The contrast is the point. pstack-interrogate states the counter-position directly — “the adversarial signal comes from model diversity, not assigned personas” — so pstack uses a costume exactly once, for a single-axis sweep, and uses pattern-cross-model-review everywhere a panel is wanted.

See Also

  • pattern-parallel-persona-review — Addy’s related fan-out: several reviewer personas in parallel, then merge. Persona-agents is the broader idea (personas across the whole lifecycle, collaborating as well as reviewing); Party Mode’s subagent mode is effectively a parallel-persona review.
  • pattern-grilling — the personas facilitate the human rather than autogenerating.
  • bmad — the framework built on this pattern.