Techcle Wiki

SKILL

3 min read

BMAD Party Mode

Source material: /workspace/_bmad/core/workflows/party-mode/ (workflow.md + steps/)

This skill adapts the BMAD party-mode workflow for single-agent Paperclip heartbeats. Rather than spawning multiple live agents, the invoking agent voices 2–3 role archetypes in turn and synthesises their exchange. The invoking agent’s own identity is never overridden — archetypes are lenses, not personas.

Role Archetypes

ArchetypeFocusTypical concerns
platform-engInfra, reliability, scalabilityOps burden, SLAs, runbook coverage, cost
full-stackImplementation feasibilityAPI design, DX, test coverage, edge cases
devsecopsSecurity & complianceAttack surface, secrets, least-privilege, audit trail
ctoTechnical strategyArchitecture fit, tech debt, build-vs-buy
ceoBusiness impactScope, timeline, revenue, stakeholder risk
poDelivery & qualityAcceptance criteria, Definition of Done, prioritisation

When to Invoke

  • A ticket has ambiguous scope or competing approaches.
  • A decision has non-obvious cross-functional implications.
  • An IC or manager wants a structured challenge before committing to a plan.
  • A blocked issue needs fresh perspectives to find an unblocking path.

Execution Steps

1. Load Issue Context

Read the issue being discussed. Gather:

  • Title, description, and acceptance criteria.
  • Current status, any plan document, and recent comments.
  • Parent/ancestor context if relevant.

2. Select Archetypes

Choose 2–3 archetypes whose concerns are most relevant to the issue:

  • Match primary archetype to the dominant domain (e.g. security issue → devsecops first).
  • Add a complementary perspective that might challenge or enrich the first.
  • Optionally add a third if a strategic or delivery dimension is genuinely distinct.
  • Rotate archetypes across sessions to avoid repetitive groupings.

3. Facilitate the Exchange

For each selected archetype, produce a short in-character response:

**[archetype-name]:** [Concise, focused point — 2–5 sentences]

Archetypes may address each other directly:

“As full-stack noted, the API boundary is the tricky part — but from a devsecops angle, the bigger risk is…”

Keep each archetype’s voice consistent with its focus area (see table above). Allow respectful disagreement. Avoid circular repetition — if a point has been made, build on it or challenge it rather than restating it.

4. Synthesise

After the exchange, produce a Structured Summary section:

## Party Mode Summary
 
### Insights
- [key finding 1]
- [key finding 2]
 
### Agreements
- [shared conclusion 1]
 
### Open Questions
- [unresolved question 1 — tag the archetype best placed to answer]

Output Rules

  • All output goes into Paperclip only — as an issue comment or a named issue document (e.g. key party-mode-summary).
  • No BMAD metadata files, no frontmatter state files.
  • Do not claim to be any archetype as your actual identity — frame output clearly as a structured analysis exercise.
  • Keep the full exchange under ~400 words unless the issue is genuinely complex.

Identity Rule

The invoking agent speaks as each archetype for analytical purposes only. The agent’s own role, name, and Paperclip identity remain unchanged throughout. Never use archetype names in API calls, status updates, or issue assignments.

Example Invocation Pattern

When an agent invokes this skill on an issue, the output structure is:

**Invoking party mode on [CAS-NNN](...):**
 
**full-stack:** The core risk here is…
 
**devsecops:** Agreed on the implementation risk, but I'd flag that…
 
**po:** From a delivery standpoint, the acceptance criteria need to be…
 
## Party Mode Summary
### Insights
- …
### Agreements
- …
### Open Questions
- …

Post this as a comment on the issue, or write it to an issue document with key party-mode-summary if a persistent record is preferred.

Graph View