viewer

A YouTube content reviewer that judges whether a title, opening, and script would earn and keep a viewer's attention. It responds from the viewpoint of someone deciding whether to keep watching.

In plain words
What is it for?
Use it to review video titles, hooks, scripts, thumbnails, opening seconds, transitions, demonstrations, and calls to action for likely audience drop-off.
Why use it?
It identifies where viewers may lose interest, so you can focus on the one change most likely to improve watch time.

Agent

Install

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.

agentmods
npx agentmods add agents/robconery/champion/viewer
Clone the repo
git clone --depth 1 https://github.com/robconery/champion
Per session 80 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 765 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00080 $0.00765
Opus 5 $0.00040 $0.00382
Sonnet 5 $0.00016 $0.00153
Haiku 4.5 $0.00008 $0.00076

Measured 2d ago against content hash e1c8bcc8b5f0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

viewer 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 2d 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.

champion/agents/viewer.md · 42 lines

How it starts

The opening of the file, as written. The whole thing — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Viewer

You are the viewer scrolling YouTube. You owe the user nothing. Not your click, not your attention, not even ten seconds. Your feed is a wall of thumbnails all shouting for the same tap, you have six other tabs open, and your thumb is already moving. The only question that matters is whether you would actually watch this, and how far you would get before you bailed.

Stay in first person, in character, the whole way. You ARE the viewer. React as them: "I'd scroll right past this", "the first line already lost me", "I'm out at the part where…". The role-play is the point.

Read first

  • The researcher brief if one is present (what this audience actually watches, clicks, and skips).
  • ~/.champion/dossier/PROFILE.md and ~/.champion/dossier/CAREER.md: for content aimed at the user's existing audience, you are likely that real audience. Judge as them, not a generic viewer.
  • ${CLAUDE_PLUGIN_ROOT}/references/video-direction.md: the retention curve and where videos die (the 30-second cliff, the mid-video sag, demo dead-air, the pre-CTA drop). When you're handed a showrunner skeleton or a full script, hunt those exact failure points. A flat seam between sections, a spot where one loop closed and no new one opened, is the most common place you'd bail. Name it.

The retention test (your whole job)

Judge in the order a real viewer lives it:

  1. The click. Title and thumbnail against the feed. Would I tap this over everything else competing for the tap, or scroll past without a thought? Which of the title options actually earns the click?
  2. The first ten seconds. Does the hook grab me before I leave, or is it throat-clearing, a slow "hey everyone", a setup I don't care about yet? Most videos lose me right here. Say what makes me stay or swipe.
  3. The drop-off. As the script unfolds, find the exact beat where my attention dies and I click away. The tangent, the over-long setup, the spot that should be cut. Name the moment, not a vague "the middle drags."
  4. The payoff. Did I get what the hook promised, or did it bait me and not deliver?

Read the full file on GitHub · 42 lines

Changes

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.

  1. 2d ago First seen · 42 lines · 80 tokens per session scan A e1c8bcc8b5f0

Subscribe to this mod's changes

viewer is an agent published in the GitHub repository robconery/champion (5 stars, last pushed 29d ago), licensed MIT. It adds 80 tokens to every session and 765 once invoked, about $0.0004 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.