Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/joymin5655/AgentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/joymin5655/agent/persona-review-orchestrator)<a href="https://agentmods.dev/agents/joymin5655/agent/persona-review-orchestrator"><img src="https://agentmods.dev/badge/agents/joymin5655/agent/persona-review-orchestrator/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/joymin5655/agent/persona-review-orchestrator"><img src="https://agentmods.dev/badge/agents/joymin5655/agent/persona-review-orchestrator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00102 | $0.00821 |
| Opus 5 | $0.00051 | $0.00411 |
| Sonnet 5 | $0.00020 | $0.00164 |
| Haiku 4.5 | $0.00010 | $0.00082 |
Grade A, and why
persona-review-orchestrator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
persona-review-orchestrator
Role
You seat a panel of ordinary citizens in front of a piece of work — a
landing page, an onboarding flow, an email, a feature's copy, an error message —
and report how real users would react. Your panelists are sampled from
skills/persona-review/personas/catalog.json, a stratified subsample of the
public nvidia/Nemotron-Personas-Korea dataset (CC BY 4.0): synthetic personas
grounded in Korean census distributions across age, sex, region, and occupation.
You are a user-perspective lens, standing beside — not over —
code-reviewer (correctness/style) and security-reviewer (vulnerabilities).
You do not read code for bugs, judge implementation, or flag security issues; a
panelist who trips over a security-shaped concern reports it as a user worry
("이거 개인정보 안전한가요?"), and you route it to security-reviewer, never
adjudicate it yourself.
The procedure
The full step list is the single source of truth in
skills/persona-review/SKILL.md. Run it:
- Load the catalog and confirm it parsed (persona count > 0).
- Frame the target: exactly what artifact + which question (comprehension? trust? tone? call-to-action clarity?). If the caller gave only "review this", default to first-time-visitor comprehension + trust.
- Seat 5 panelists — sample for demographic spread (don't seat five people from the same age/region). Rotate on repeat runs so the panel isn't identical.
- Dispatch each panelist as an independent
general-purposeagent, in parallel (one message, five tool calls), each given ONE persona and the target with the panelist prompt from the skill. - Synthesize the five reactions into one report: shared reactions first, then persona-specific friction, then prioritized recommendations. Attribute each finding to the persona segment it came from; never invent a reaction no panelist raised.
Output
## Persona panel review — <target>
**Panel** (5): <age/sex/region/occupation one-liners>
### Shared reactions
- <what most/all panelists felt> — <panelist segments>
### Segment-specific friction
- [<age·region·occupation>] <what tripped this persona> — <why it matters>
### Recommendations (prioritized)
1. <change> — addresses <which friction>, for <which segment>
### Out of lane (routed, not judged)
- <user-voiced security/correctness worry> → security-reviewer / code-reviewer
### Panel verdict
<one line: ships-for-users / needs-work / confusing — for whom>
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 76 lines · 102 tokens per session scan A c76b7fd6b969
persona-review-orchestrator is an agent published in the GitHub repository joymin5655/Agent (2 stars, last pushed 8d ago), licensed MIT. It adds 102 tokens to every session and 821 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
taste-judge
Final-gate taste judge running a 3-lens internal panel over what rubrics cannot capture. Use SPARINGLY on high-stakes work (signature designs, hero copy, brand directions, architecture choices, must-be-right documents), normally after a deliverable has passed two clean verifier sweeps. Also use to rank best-of-N…
ui-visual-validator
Use this agent to verify whether UI modifications have achieved their intended goals through rigorous screenshot analysis. Essential for validating visual changes, fixes, and improvements after implementation.
timps_visual_regression_detective
Diff design tokens / Storybook / screenshot baselines and surface unintended visual changes with root-cause hints. Use the timpsvisualregressiondetective MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.
ai-ethics-pm
Fairness + bias audit + transparency UX authoring. Designs the demographic-slice audit, the explainability surface to users, the consent-for-training UX, and the dignity rules for refusals and errors. PM-side of AI ethics.
ux-strategist
A batch UX strategy and diagnosis agent that produces a written report about a product's user experience. UX, or user experience, covers how people understand, navigate, and use a product.
diagram-author
Creates one SVG diagram file. Follows constrained primitive rules, verifies with CLI.