Persona Review

SkillDocs & knowledge

Review a document from multiple stakeholder perspectives (personas), collect structured feedback, and optionally update the document. Use when a document should be stress-tested from different viewpoints before sharing — e.g., "review this PRD from an engineer's perspective", "what would a skeptical investor say about this pitch?" Triggers on persona review, review from a stakeholder perspective, what would an engineer say, stress-test this PRD, review this pitch, skeptical investor.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Persona Review skill

What this skill tells your AI

The instructions your AI receives, as published by glebis/humane-agentic-design in humane/skills/persona-review/SKILL.md and read by ahel’s review.

Announce at start: "I'm using the humane:persona-review skill to stress-test this document from stakeholder perspectives."

Overview

Takes a document and reviews it from N configurable personas. Each persona reads the document independently, provides structured feedback, then optionally the document is updated to address the feedback.

Usage

/persona-review [file-path] [personas]

Arguments:

  • file-path — path to the document to review (required). If omitted, ask user.
  • personas — comma-separated persona names or "auto" (default: auto)

Examples:

/persona-review PRD.md engineer,designer,exec
/persona-review "Claude-Drafts/pitch-deck-script.md" auto
/persona-review  # will ask for file and use auto personas

Workflow

Step 1: Load Document

Read the target file. Extract its purpose from context (frontmatter, filename, content).

Step 2: Select Personas

If auto: Choose 3 personas most relevant to the document type:

  • PRD/spec → Skeptical Engineer, UX Designer, Business Stakeholder
  • Pitch/proposal → Potential Customer, Investor, Competitor
  • Blog post/article → Target Reader, Editor, Subject Expert
  • Teaching material → Beginner Student, Advanced Practitioner, Curriculum Designer
  • Marketing copy → Target Audience Member, Brand Skeptic, SEO Specialist

If specified: Use the named personas. Interpret flexibly — "engineer" becomes "Senior Engineer who has seen too many half-baked specs."

Step 3: Review Rounds

For each persona, generate a review with this structure:

### [Persona Name]
**Perspective:** [1-line description of who they are and what they care about]

**Reaction:** [2-3 sentences — their gut reaction on first read]

**Strengths:**
- [What works well from this perspective]

**Concerns:**
- [What's missing, unclear, or problematic]

**Key Question:** [The single most important question this persona would ask]

**Suggestion:** [One specific, actionable improvement]

Step 4: Synthesis

After all persona reviews, create a synthesis section:

## Synthesis

**Consensus strengths:** [What all personas agreed works]
**Consensus gaps:** [What multiple personas flagged]
**Tensions:** [Where personas disagree — these are the interesting trade-offs]
**Priority changes:** [Top 3 changes ranked by impact]

Step 5: Ask User

Present the full review and ask:

"Apply the top changes to the document? (all / pick / none)"

  • all — Apply all priority changes
  • pick — Let user select which to apply
  • none — Leave document unchanged, keep review as reference

Step 6: Apply (if requested)

Edit the original document to address selected feedback. Keep changes minimal and targeted. After editing, show a brief diff summary of what changed.

Step 7: Save Review

Save the review alongside the document:

  • If document is path/to/doc.md, save review as path/to/doc-persona-review.md
  • Add frontmatter linking back to the original document:
---
type: persona-review
source: "[[original-document]]"
personas: [list]
created_date: '[[YYYYMMDD]]'
---

Guidelines

  • Each persona should have a distinct voice — don't make them all sound like the same reviewer with different labels
  • Personas should be constructively critical, not adversarial
  • Focus on substance (logic, completeness, clarity) not style (grammar, formatting)
  • If the document is in Russian, reviews should be in Russian
  • Keep each persona review under 200 words — brevity forces prioritization
  • The "Key Question" is the most valuable output — it should be something the author genuinely hadn't considered

Signals

GitHub stars
28
Forks
1
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
persona-review-glebis
Source
github.com/glebis/humane-agentic-design