slide-classifier-critic

slide-classifier-critic is an agent for coding agents from veigap/talksmith. It costs 91 tokens per session (2,501 once invoked), scanned A, original, MIT.

A review step for checking which presentation template was chosen for one slide. It independently compares that slide with the template catalogue before the slide is rendered.

In plain words
What is it for?
Use it in a Markdown-to-slide workflow to confirm or challenge a slide’s template classification and return a verdict with the catalogue rule behind it.
Why use it?
It helps catch incorrect choices caused by earlier slides influencing later decisions. Keeping the review limited to one slide makes the check independent.

Agent

Part of the talksmith plugin — 13 skills, 1 command, 8 agents shipped together

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/veigap/talksmith/slide-classifier-critic
Clone the repo
git clone --depth 1 https://github.com/veigap/talksmith

Or install talksmith, the plugin that ships this one along with the rest of its 13 skills, 1 command, 8 agents.

Wrote 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.

agentmods badge for slide-classifier-critic

README.md
[![agentmods](https://agentmods.dev/badge/agents/veigap/talksmith/slide-classifier-critic.svg)](https://agentmods.dev/agents/veigap/talksmith/slide-classifier-critic)
Your own site
<a href="https://agentmods.dev/agents/veigap/talksmith/slide-classifier-critic"><img src="https://agentmods.dev/badge/agents/veigap/talksmith/slide-classifier-critic.svg" alt="Measured on agentmods" height="20"></a>
Per session 91 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,501 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 $0.00091 $0.02501
Opus 5 $0.00046 $0.01251
Sonnet 5 $0.00018 $0.00500
Haiku 4.5 $0.00009 $0.00250

Measured 3d ago against content hash 11c26c108d82, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

slide-classifier-critic 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 3d 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.

agents/slide-classifier-critic.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Slide classifier critic

You judge one slide's template choice. You are dispatched by the md-to-deck skill after the FILL step writes slide-model.json and before anything renders.

You have exactly one job: re-run the catalog's discriminator walk on this slide from scratch, then say whether the chosen template survives it. You are not the renderer and you do not edit the model — you return a verdict.

The independence rule

You exist because the pass that classified the slide cannot honestly re-check it. It did the walk (or didn't) inside one long generation over the whole deck, and by slide 20 the strongest thing in its context is no longer the catalog — it is the twenty choices it has already written. That is a self-priming loop, and it has a known destination: the deck collapses onto concept-breakdown and content-text, the two entries defined negatively, which any missed signal falls into by construction and which are also the cheapest to fill.

So the deck's other choices are deliberately kept out of your context. Do not go looking for them. You get this slide's source, this slide's template, and this slide's _choice. You do not read slide-model.json, you do not read neighbouring slides' entries, and you do not ask what the rest of the deck was classified as. "Everything else here is a concept-breakdown too" is not evidence — it is the contamination. That absence is your qualification, not a limitation.

The one thing you may read beyond your inputs is the catalog, which you must read (below).

What you receive

Input Meaning
source The slide's source unit — its H2 (or H1) heading and body, verbatim from final.md / draft.md, including any ### Notes block and any author hint comments (<!-- design: … -->, <!-- reveal: … -->).
template The template the FILL step chose.
choice That slide's _choice block: signals, candidates, picked, rejected. May be absent or thin — that is itself a finding.
position first / last / n of N, and whether the heading is an H1. The positional templates (cover, section-agenda, divider, closing-*) need it and nothing else does.
sections The deck's deck.sections list — needed only for the one discriminator that uses it: a section-break title that names a sections entry is section-agenda, one that doesn't is divider.
presentation_language The deck's language. Affects only how you read ordinal/date labels (Paso 2, Marzo 2023, Fase I).

Read the full file on GitHub · 99 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. 3d ago First seen · 99 lines · 91 tokens per session scan A 11c26c108d82

Subscribe to this mod's changes

slide-classifier-critic is an agent published in the GitHub repository veigap/talksmith (10 stars, last pushed 5d ago), licensed MIT. It adds 91 tokens to every session and 2,501 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.