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 skills add Syysean/web-ai-skills --skill deepseek-workergit clone --depth 1 https://github.com/Syysean/web-ai-skillsWrote 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/skills/syysean/web-ai-skills/deepseek-worker)<a href="https://agentmods.dev/skills/syysean/web-ai-skills/deepseek-worker"><img src="https://agentmods.dev/badge/skills/syysean/web-ai-skills/deepseek-worker/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/skills/syysean/web-ai-skills/deepseek-worker"><img src="https://agentmods.dev/badge/skills/syysean/web-ai-skills/deepseek-worker.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 | $0.00077 | $0.01401 |
| Opus 5 | $0.00039 | $0.00700 |
| Sonnet 5 | $0.00015 | $0.00280 |
| Haiku 4.5 | $0.00008 | $0.00140 |
Grade C, and why
deepseek-worker scanned grade C with 1 finding 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 11d 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.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
``` Verify returned value matches the intended prompt. If mismatch: retry. Take a screenshot to confirm text appears in the textarea. --- ### Step 5 — Send DeepSeek's send button is a div with role="button", NOT a nativ How it starts
The opening of the file, as written. The whole thing — 216 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DeepSeek Worker
Send a prompt to DeepSeek via browser and return the response.
Environment
- URL: https://chat.deepseek.com
- Browser: Chrome on localhost:9222 (managed by local-browser-tools)
- Login: Required — session persisted in ~/.cache/browser-tools profile
- Tools:
browser-start.js,browser-nav.js,browser-eval.js,browser-screenshot.js
Input
prompt: string # The text to send to DeepSeek
task_type: string # qa | summary | translation | rewrite | extraction | comparison | brainstorming | analysis
Output
Always output this JSON block first:
{
"site": "deepseek",
"task_type": "<task_type>",
"status": "success | partial | error",
"answer": "<extracted response text>",
"retry_count": 0,
"elapsed_ms": 0,
"error_reason": null
}
Then follow with a short human-readable summary (2–3 sentences max).
Workflow
Step 1 — Start browser
node ~/.pi/agent/skills/local-browser-tools/browser-start.js
Expected: ✓ Chrome already running on :9222 or ✓ Chrome started on :9222
If fails: return error status, stop.
Step 2 — Navigate to DeepSeek
node ~/.pi/agent/skills/local-browser-tools/browser-nav.js https://chat.deepseek.com
Take a screenshot and verify:
- Textarea input box is visible with placeholder "Message DeepSeek"
- Not showing a login form
If login page appears: return error_reason: "login_required", stop.
Step 3 — Locate input box
DeepSeek uses a <textarea> with placeholder="Message DeepSeek". Note: no aria-label.
Check:
!!document.querySelector('textarea[placeholder="Message DeepSeek"]')
Fallback selectors in order:
textarea[placeholder="Message DeepSeek"]textarea
If no selector works after 3 attempts: trigger retry.
Step 4 — Insert prompt
DeepSeek uses React — setting textarea.value directly does NOT update React's internal state and will cause "Message is empty" on send. Must use nativeInputValueSetter:
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.
- 11d ago First seen · 216 lines · 77 tokens per session scan C 7b19aa08a948
deepseek-worker is a skill published in the GitHub repository Syysean/web-ai-skills (5 stars, last pushed 3mo ago), licensed MIT. It adds 77 tokens to every session and 1,401 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
edgeone skill scanner
A local static scanner that checks agent-skill files for security risks before they are installed or used. Static analysis examines files without running them.
sandbase
Access 2,000+ AI models and API tools through one MCP interface for inference, media generation, search, scraping, embeddings, social data, and structured retrieval. Use sandbasediscover before building custom integrations or declaring external data inaccessible; prefer an existing dedicated tool or API key when the…
dev-workflow
The complete development workflow for SkillHub contributors including local dev, staging validation, testing, and PR creation. Ensures agents follow the correct sequence of steps.
video-frames
Extract a single frame from a local video at the first frame, a timestamp, or a zero-based frame index using FFmpeg.
convex-performance-audit
Audits Convex performance for reads, subscriptions, write contention, and function limits. Use for slow features, insights findings, OCC conflicts, or read amplification.
convex-insights
Query a running Convex app's logs + health in natural language (official MCP): failures, slow/expensive functions, deploy causality — scoped, evidence-backed, with a dashboard deep link.