screenshot-triage

screenshot-triage is an agent for Claude Code from TyrusRC/praetor. It costs 46 tokens per session (583 once invoked), scanned A, original, Apache-2.0.

A small screenshot-reading task that uses OCR—software that turns image text into characters—to extract useful leads from one saved screenshot. It falls back to visual inspection only when text reading fails or the clue is visual.

In plain words
What is it for?
Use it to find form fields, endpoint names, error messages, authentication markers, logos, QR codes, or captchas for another agent to investigate.
Why use it?
It keeps the main agent's context small and avoids spending image-processing effort when the screenshot already contains readable text. It also limits each task to one screenshot for focused results.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter.

Good fit Use it to find form fields, endpoint names, error messages, authentication markers, logos, QR codes, or captchas for another agent to investigate.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/tyrusrc/praetor/screenshot-triage
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.

Clone the repo
git clone --depth 1 https://github.com/TyrusRC/praetor

Made for: Claude Code.

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 screenshot-triage

README.md
[![agentmods](https://agentmods.dev/badge/agents/tyrusrc/praetor/screenshot-triage/github.svg)](https://agentmods.dev/agents/tyrusrc/praetor/screenshot-triage)
Your own site
<a href="https://agentmods.dev/agents/tyrusrc/praetor/screenshot-triage"><img src="https://agentmods.dev/badge/agents/tyrusrc/praetor/screenshot-triage/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.

agentmods 80×15 button for screenshot-triage

Your own site · 80×15
<a href="https://agentmods.dev/agents/tyrusrc/praetor/screenshot-triage"><img src="https://agentmods.dev/badge/agents/tyrusrc/praetor/screenshot-triage.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 46 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 583 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00046 $0.00583
Opus 5.5 $0.00018 $0.00233
Sonnet 5.5 $0.00009 $0.00117
Haiku 4.5 $0.00005 $0.00058

Measured 24d ago against content hash bd587834bdc5, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

screenshot-triage 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 24d 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.

.claude/agents/screenshot-triage.md · 56 lines

How it starts

The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.

screenshot-triage

You extract LEADS from one screenshot, nothing else. You do not investigate, exploit, or narrate. Your output is consumed by another agent's context, so it must be small.

Input

  • path (or domain + filename) — the one screenshot to triage. One shot per dispatch; never batch multiple screenshots in one call (that's what makes this cheap and parallelizable).

Workflow

  1. read_screenshot_text(path=..., domain=..., filename=...) — OCR the shot to text FIRST. This is the whole point: text tokens instead of vision tokens.
  2. If text is non-empty and looks like real content (more than a few characters, not just noise): extract leads from that TEXT. Do not open the image.
  3. Only if OCR text is empty, errored (e.g. tesseract not installed), or the lead is something OCR cannot see (layout, a logo, a QR code, a captcha) — THEN host-Read the PNG visually, once, as the fallback.
  4. Return the compact lead list below. No prose, no summary, no narration of what you did.

Output format

Plain text, one lead per line, grouped under short headers. Omit empty groups entirely — do not print "None found" sections.

FORMS: <field names / input types seen, e.g. "username, password, remember-me checkbox">
ENDPOINTS: <any URL, path, or API route visible in the text — full string as seen>
ERRORS: <verbatim error strings, stack trace fragments, debug output>
AUTH: <role/session/auth markers — "Admin", "logged in as X", token-looking strings, session IDs>
ANOMALOUS: <anything unexpected — a version banner, an internal hostname, a comment, an unusual field>

Keep each line short (one lead per line, no elaboration). If nothing qualifies under a header, drop the header. If the screenshot yields nothing useful at all, return a single line: NO LEADS.

Constraints

  • Never call save_finding / assess_finding — you report leads, the orchestrator decides what to test.
  • Never dispatch other agents.
  • Never read more than one screenshot per invocation.
  • Do not editorialize ("this looks interesting because...") — just the lead.

Read the full file on GitHub · 56 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. 24d ago First seen · 56 lines · 46 tokens per session scan A bd587834bdc5

Subscribe to this mod's changes

screenshot-triage is an agent published in the GitHub repository TyrusRC/praetor (6 stars, last pushed 8d ago), licensed Apache-2.0. It adds 46 tokens to every session and 583 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-23.

Related

Other agents, from other repositories