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.
npx agentmods add skills/agentcomputerai/torch/scrapenpx skills add AgentComputerAI/torch --skill scrapegit clone --depth 1 https://github.com/AgentComputerAI/torchWhat 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 | $0.00041 | $0.03464 |
| Opus 5 | $0.00020 | $0.01732 |
| Sonnet 5 | $0.00008 | $0.00693 |
| Haiku 4.5 | $0.00004 | $0.00346 |
Grade A, and why
scrape scanned grade A with 2 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 yesterday.
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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- Phase 0 (curl): 2 minutes max Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
import { spawn } from "node:child_process"; How it starts
The opening of the file, as written. The whole thing — 289 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Goal of a first-time scrape
You are a scout, not a crawler. On a site that does not already have a skill in skills/sites/<slug>/, the goal is to prove a strategy works on a small sample (5-20 items) and write a reusable playbook — NOT to extract every row on the site. Full crawls come on re-runs once the skill exists.
Budget
- ~5 minutes total wall clock for a new-site scrape. The user is watching a clock you cannot see.
- Per-phase time boxes:
- Phase 0 (curl): 2 minutes max
- Phase 1 (framework recon): 3 minutes max
- Phase 2 (browser): 5 minutes max
- Once extraction works on even 5 valid items, STOP. Commit to that approach, write the skill, ship. Do not switch strategies to find a cleaner one. Do not chase pagination unless explicitly asked. Every extra minute you spend "improving" a working scrape is a minute spent not writing the skill.
- After two failed strategies on the same page, stop escalating. Write a skill documenting what you tried and which layer failed. A half-skill is more valuable than a timeout with nothing saved.
Background execution
Run scrapers as background processes, not synchronous bash. You have pi-processes tools — use them. Synchronous bash node scrape.js blocks the turn, makes you blind to progress, and prevents steering. Background execution lets you watch logs, catch errors early, and kill a run once you have enough data.
Pattern:
- Write
scrape.js - Spawn it as a background process, tag it with the slug
- Tail its log output
- As soon as you see the first 5 items extract cleanly, you can either let it finish or kill it
- Read
./output/<slug>.json, verify, write the skill
Naming invariants
Derive the slug ONCE at the start of the scrape from the brand/site name. Lowercase [a-z0-9-] only, no dots, no TLDs, no www. prefix. Examples: amazon, hackernews, mcmaster, digikey.
- Output file:
./output/<slug>.json— one file per scrape, no variants - Skill file:
./skills/sites/<slug>/SKILL.md— directory name === slug === frontmattername: - Scratch files during exploration: prefix with
_scratch_so they're easy to delete
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- yesterday First seen · 289 lines · 41 tokens per session scan A 6943e63156ec
scrape is a skill published in the GitHub repository AgentComputerAI/torch (5 stars, last pushed 4mo ago), licensed MIT. It adds 41 tokens to every session and 3,464 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 2 findings (makes network calls, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
tiny-web-crawler
Crawl from one or more starting web pages, fetch readable content, search within pages, follow relevant links, and stop when the requested information is found or a bounded limit is reached.
playwright-cli
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
browserstack
../../../engineering-team/playwright-pro/skills/browserstack/SKILL.md.
analyze-performance-traces
Analyze Chrome, Chromium, Electron, React DevTools, or Perfetto-compatible JSON traces and audit user-reported profiling findings without loading large artifacts into context; prove trigger-to-render/layout chains, separate measured facts from source inference, find exact code choke points, classify forced layout and…
feature-demo-recording
Record a demo video of a web feature from a real browser. Two modes -- a NARRATED film where measured voiceover drives the timeline (designed slides, subtitles, punch-in camera, rendered from an HTML timeline), and a SILENT evidence clip for a PR or a QA pass. Use when the user asks to record a video, demo, or screen…
browser-recording
Record a browser flow as a video/GIF for evidence — animations, transitions, and multi-step interactions that a still screenshot cannot prove. Drives the project's own Playwright through a bundled runner, then converts to mp4 + GIF via ffmpeg. Use when the user asks to record a demo, capture a GIF or video of the UI…