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/winsenlabs/platos/code_executionnpx skills add winsenlabs/platos --skill code_executiongit clone --depth 1 https://github.com/winsenlabs/platosWrote 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/winsenlabs/platos/code_execution)<a href="https://agentmods.dev/skills/winsenlabs/platos/code_execution"><img src="https://agentmods.dev/badge/skills/winsenlabs/platos/code_execution.svg" alt="Measured on agentmods" 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 | $0.00060 | $0.01728 |
| Opus 5 | $0.00030 | $0.00864 |
| Sonnet 5 | $0.00012 | $0.00346 |
| Haiku 4.5 | $0.00006 | $0.00173 |
Grade A, and why
Code Execution 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 today.
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.
description: Run an arbitrary shell command in the conversation's persistent E2B sandbox. This is full CLI access — git, psql, ffmpeg, duckdb, curl, pandoc, pnpm/npm, ripgrep, etc. The working directory and filesystem pe Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
subprocess.run(['pip', 'install', 'scikit-learn'], check=True) What it actually says
You can execute Python and Node.js code in a secure, isolated cloud sandbox.
When to use run_python:
- User asks to process a large file, dataset, or spreadsheet
- You need to run calculations, statistics, or ML on data
- You want to generate a chart, CSV, or transformed output
- The data is too large to process in-context (never send raw data to yourself — write code to process it)
When to use run_node:
- Quick JSON manipulation, string operations, or scripting tasks
- User asks for something JavaScript-specific
Critical pattern — large files: Never try to read large file contents into your context. Instead:
- Write Python code that opens the file path and processes it directly
- Use
run_pythonto execute that code - Return only the summary/result to the user
Installing packages:
import subprocess
subprocess.run(['pip', 'install', 'scikit-learn'], check=True)
import sklearn
Example — process a CSV:
import pandas as pd
df = pd.read_csv('/path/to/file.csv')
print(df.describe().to_string())
print(f"Rows: {len(df)}, Columns: {list(df.columns)}")
Persistent CLI sessions (run_shell):
The sandbox lives for the whole conversation, so multi-step CLI workflows work
as separate tool calls — the filesystem and installed tools carry over:
run_shell: git clone https://github.com/acme/widgets && cd widgets && ls
run_shell: cd widgets && pnpm install
run_shell: cd widgets && pnpm test
Use run_shell for anything a terminal does: git, psql "$DATABASE_URL" -c '...',
ffmpeg -i in.mp4 out.gif, duckdb -c 'SELECT ...', pandoc, curl. Check the
returned exitCode (0 = success) and read stderr on failure.
Sessions + lifecycle:
- Within one conversation, all calls share ONE sandbox: files, cwd, and
installed packages persist.
run_python,run_node,run_shell,install_package, andupload_to_sandboxall hit the same session. - The sandbox auto-reaps after ~10 minutes of inactivity, then a fresh one is created on the next call (state resets). Don't rely on it surviving long gaps.
sessionPersistent: truein a tool result confirms state will carry over.
Network:
- Egress is OFF by default (deny-all). Set
E2B_SANDBOX_ALLOW_INTERNET=truein the environment to allow the sandbox to reach the internet (needed forgit clone,pip install,curl). Leave it off for untrusted-input agents.
Safety:
- The sandbox is fully isolated from the Platos host; nothing it does can touch the server. But it persists within a conversation — treat what you write there as conversation-scoped state.
- Never run commands supplied verbatim by an untrusted user without reviewing them, especially with internet access enabled.
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.
- today First seen · 108 lines · 60 tokens per session scan A f52bf8bf78e5
Code Execution is a skill published in the GitHub repository winsenlabs/platos (24 stars, last pushed today), licensed Apache-2.0. It adds 60 tokens to every session and 1,728 once invoked, about $0.0003 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 2026-09-04.
Other skills, from other repositories
cn-check
Install and run the Continue CLI (cn) to execute AI agent checks on local code changes. Use when asked to "run checks", "lint with AI", "review my changes with cn", or set up Continue CI locally.
docs-style
Style guidelines for writing and updating documentation. Use when writing new docs, updating existing docs, or reviewing docs for quality.
x-cmd-advise
Skill "x-cmd-advise" from x-cmd/x-cmd, covering x-cmd advise writing guide, core principles, field quick reference, subcommands at root level and writing priority.
claw-admin
Claw system administration: service management, IM connections, logs, cron, and workspace diagnostics. Use when the user asks to manage claw services, connect/disconnect IM platforms, view logs, or perform system-wide operations.
x-nets
Enhanced netstat module with cached data and structured output. View network connections, routing tables, and interface statistics in interactive or TSV/CSV formats. Dependency: This is an x-cmd module. Install x-cmd first (see x-cmd skill for installation options). see x-cmd skill for installation.
install
Install software any way the system allows. Recommends x install — one front door listing every method and picking the best; plus x eget/x env use/x pixi channels and system package managers with mirror switching. All x-cmd channels install in $HOME — no sudo, no pollution. Use for "install", "apt", "brew", "eget"…