Skip to content

forcedotcom/sf-skills / plugins/builder/agentforce-adlc/skills/agentforce-persona-generate

agentforce-persona-generate

Verified

Use to design an AI agent persona — identity, voice, tone, behavioral style, guardrails — and encode it into Agent Script (.agent files) or Agentforce Builder field values. Runs four sequenced phases (DRAFT, REFINE, SCORE, ENCODE) plus five spokes (NAME, FILL OUT, ADJUST, CHECKLIST, WORKSHOP). Supports greenfield (brand input to a starter .agent file) and brownfield (audit an existing .agent file and re-encode persona fields only), with a 100-point evaluation rubric. Trigger when a user designs a new agent persona, retrofits persona consistency, translates brand guidelines into agent voice, defines tone/register/formality/warmth/humor/personification, encodes a persona for Agent Script or Agentforce Builder, or scores or audits an existing .agent file for persona quality. Do not trigger for building Agentforce actions/metadata, authoring Agent Script flow control without persona work, or testing agent conversations.

references/workshop-question-policy.md

View on GitHub
---
version: "1.0.0"
date: 2026-05-15
---

# Workshop Question Policy

Rules for generating stakeholder-facing questions in the WORKSHOP spoke. The skill produces these questions for a customer's internal stakeholders to answer **asynchronously** — not in a live workshop with the skill running. So the questions need to be open-ended, thought-provoking, scoped to what's actually unresolved, and respectful of stakeholder time.

## Guiding principle

**Quality over coverage. Better to ask 4 great questions than 12 generic ones.**

There is an infinite possible question set the framework could generate. Most of those questions are noise for any specific persona. The job here is identifying *what's actually unresolved or high-leverage* for *this* persona, given everything the skill already knows.

## Skip what you already know

The skill must NOT generate questions for things it already knows from the brand input, persona draft, or context. Stakeholder attention is the scarce resource.

**Examples of what to skip:**

- The brand input clearly says "always formal" → don't ask about Formality.
- The customer is a serious law firm → don't ask "How much humor?" (the answer is probably None and asking burns attention on a non-question).
- Channel is locked to telephony → don't ask "Should the agent use emoji?" (no formatting in audio).
- Personification is locked to Talking System → don't ask about character names.
- Identity is filled with strong-signal traits → don't ask the stakeholders to re-validate every adjective; pick one or two with productive tension.

This is harder than it sounds. The framework has 14 dimensions × 5+ persona artifacts × business context. Most could generate plausible questions. The skill must actively prune.

## What to ask about

In priority order:

1. **High up the dimensions ladder** — Identity, Personification, Register. These have the most leverage; getting them wrong cascades.
2. **Where confidence is low** — when a draft dimension was filled with a default rather than a strong signal, that's a candidate question.
3. **Where productive tension exists** — Tension Pairs in the framework (Cool Warmth + Bold Personality, Blunt Coloring + Playful Humor, etc.) are inherently questions stakeholders should weigh in on.
4. **Where the brand input is silent or contradictory** — e.g., the brand guide says "approachable" but doesn't say how much humor; that's worth asking.
5. **Where channel forces a tradeoff** — voice agents handling frustrated callers; humor in escalation flows.
6. **Where the use case raises a tone-flex question** — the framework's tone-flex triggers (system errors, content sensitivity, escalation) are stakeholder-relevant moments.

## What never to ask about

- Things the input or context already answers definitively.
- Things downstream of a higher decision already made (don't ask about emoji if Formal Register is locked — Formal pulls Emoji to None).
- Internal mechanics the stakeholders shouldn't be making decisions about (e.g., dimension constraint notes, tension-pair resolution rules).
- Anything that produces a "neutral default" answer — those questions waste stakeholder attention without changing the persona.

## Question styles

Three styles, picked per question based on what's needed:

### 1. Validate

Use when the skill has a strong-signal draft and wants stakeholder confirmation.

> "We assumed [X] from your brand guide. Confirm or push back."
> "The agent will speak in Coach register — instructive, takes a position. That fits the 'sales coach' framing in the brief. Right call?"

### 2. Choose between

Use when there are two genuinely different positions on a spectrum, with consequences spelled out.

> "When a customer expresses frustration, would you rather the agent (A) acknowledge the frustration before solving, or (B) skip to the solve? A reads as more empathetic; B reads as more efficient. Pick one."

The two options must be **meaningfully different**. Don't write false binaries where both options are the same with cosmetic differences. If you can't articulate the consequence of each choice, drop the question.

### 3. Open spectrum

Use when it's a blank slate — no signal in the input, and the stakeholders genuinely need to make the call.

> "Place the agent on this line: [Brisk] ⟷ [Languid]"
> "How much character should come through? [Talking System] ⟷ [Familiar Thing] ⟷ [Personal Assistant]"

These are dimension-card-shaped — print on a page, dot-vote, or place stickies on a FigJam board. Especially good for high-leverage dimensions (Identity, Personification, Register, Warmth) where the stakeholder team needs to converge.

## Output format constraints

The workshop outline is meant to be **lightweight and printable** — a FigJam board, a whiteboard, a Google Doc with sticky notes. Not a slide deck. Not a long-form persona document. The skill's job is to produce the outline; the customer's job is to run the workshop with it.

- Aim for 4-8 total questions per workshop outline. Fewer if the persona is well-grounded.
- Group questions by Activity (Identity calibration, Tone in difficult moments, Spectrum check) so the workshop has a clear arc.
- Pre-read materials section: link to or attach the brand-grounded persona draft and sample dialog so stakeholders walk in informed.
- Include a brief "Background" paragraph for the facilitator (≤1 paragraph) summarizing where the persona stands and what input is needed.

## Anti-patterns to refuse

- A 20-question workshop outline. If you're hitting 20, you haven't pruned.
- A workshop outline for a Talking System that asks "What kind of humor?" Skip. You know the answer.
- A workshop outline that asks every dimension "where on the spectrum?" — that's a coverage approach, not a leverage approach.
- A workshop outline where every option in a "choose between" pair reads as neutral or compromised. The job is to surface real tradeoffs, not to write safe choices.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall