AWorld is an agent harness, meaning a framework that coordinates an AI agent’s tools, memory, context, and execution so expert knowledge can be turned into reusable skills and autonomous agents. It is for building domain-specific agent applications and workflows, with the catalogue entries representing skills, agents, and commands that operate within the AWorld ecosystem.
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 inclusionAI/AWorld --skill filexgit clone --depth 1 https://github.com/inclusionAI/AWorldWrote 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/inclusionai/aworld/filex)<a href="https://agentmods.dev/skills/inclusionai/aworld/filex"><img src="https://agentmods.dev/badge/skills/inclusionai/aworld/filex/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/inclusionai/aworld/filex"><img src="https://agentmods.dev/badge/skills/inclusionai/aworld/filex.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00087 | $0.05498 |
| Opus 5.5 | $0.00035 | $0.02199 |
| Sonnet 5.5 | $0.00017 | $0.01100 |
| Haiku 4.5 | $0.00009 | $0.00550 |
Grade B, and why
filex scanned grade B 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 12d 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
chmod 600 /workspace/filex-options.json How it starts
The opening of the file, as written. The whole thing — 516 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Use FileX
FileX is disabled by default. Choose this skill when the task needs content from an existing document or media source. FileX does not need to run for unrelated tasks or to create new documents or media.
When the caller explicitly selects this skill, the selection is an execution
requirement rather than a suggestion. Complete the FileX workflow before using a
generic parser or claiming the task is finished. In a managed artifact-producing
runtime, create the canonical bundle under FILEX_ARTIFACTS_ROOT and require a
successful wrapper result. If FileX cannot parse the supplied source, report the
FileX failure; do not silently fall back to another parser or synthesize evidence.
An unselected baseline run must not invoke FileX merely because the built-in skill
is installed.
Use the bundled wrapper for FileX inspect, parse, and status. It validates workspace paths, resolves supported URL sources, keeps credentials out of command-line arguments, preserves FileX JSON fields, and returns output_path for synchronous parsing.
Use the AWorld runtime contract
Invoke FileX through the bundled AWorld wrapper from any working directory:
python3 /skills/filex/scripts/filex.py --help
Do not change into /app/mcp_servers/filesystem_server, call ./bin/filex, or
use /root/fs_workspace. Those paths belong to a different MCP filesystem-server
image. In a managed AWorld sandbox, FILEX_WORKSPACE_ROOT is commonly
/workspace; standalone AWorld images may use $HOME/workspace. Use the
runtime-provided value and the task's supplied path instead of deriving a path
from the current directory. /skills/filex is read-only and is never an output
location.
The wrapper accepts local --input files under FILEX_WORKSPACE_ROOT and
HTTP(S) --url sources. Keep the skill focused on those ordinary AWorld source
flows rather than deployment-specific object-store workflows.
Follow the parsing workflow
- Identify the source and confirm that a local file already resides under
FILEX_WORKSPACE_ROOT; otherwise use a trusted HTTP(S) URL or ask the caller to stage it. Do not make an unnecessary base64 copy of a mounted file. - Let FileX infer the file type when the suffix and content are reliable. Use
--file-typeonly to resolve ambiguity. - Let the configured provider handle ordinary files. Select an explicit provider only when the task needs a known capability such as scanned-page OCR, real layout geometry, or text-layer-first VLM fallback.
- Keep a simple provider choice in
--provider. Put provider-private settings and structured values in a protected workspace JSON file and use--env-file; never combine those two options. - Run the parse synchronously in the foreground with enough terminal-tool time. For evaluation or durable handoff, request an artifact bundle and an explicit layout format.
- Use the JSON result, returned paths, page coverage, and warnings to decide whether further inspection would help the task. Keep established facts in task state and revisit them when the source, artifacts, or requirements change.
- On failure, preserve the wrapper's JSON error, provider/model identity, task id, and any checkpoint. Retry only after the first process has ended and the failure is understood.
What ships with it
2 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.
- 12d ago Changed · +9 lines 2084f1bd41e9
- 1mo ago Changed · +2 lines 25d3679daa1e
- 1mo ago Changed 9c002009f7c3
- 1mo ago Changed · +391 lines scan A → B 4681e9a58d63
- 1mo ago First seen · 114 lines · 87 tokens per session scan A d398abfa3f25
filex is a skill published in the GitHub repository inclusionAI/AWorld (1,246 stars, last pushed 2d ago), licensed MIT. It adds 87 tokens to every session and 5,498 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
foundry-hosted-agent-validation
Step-by-step process for validating a Python Foundry hosted agent sample (under python/samples/04-hosting/foundry-hosted-agents/) end to end — running it locally (native runtime and azd ai agent run) and after deploying it to an Azure AI Foundry project with azd. Use this when asked to validate a hosted agent sample.
hive.pdf
Read, write, merge, split, rotate, watermark, encrypt, and OCR PDF files using Python (pypdf, pdfplumber, reportlab, pypdfium2) and command-line tools (poppler-utils, qpdf). Use when the user asks to extract text/tables/images from a PDF, create or modify a PDF, combine or split PDFs, OCR a scanned PDF…
Create and edit PDFs to a professional standard — generate from content, draw vector/precise layouts, embed custom (incl. CJK/Unicode) fonts, CREATE and fill AcroForm fields (text/checkbox/radio/dropdown), flatten, merge/split/rotate, stamp/watermark, encrypt, and extract text. Use when the user asks Brigade to make a…
pdf-processor
Extracts text and tables from PDF files, fills forms, and merges documents. Use when working with PDF files or when the user mentions PDFs, forms, or document extraction.
pdfcraft
Headless PDF work with PdfCraft, provisioned on demand by pi-craft - combine, split, extract, rotate, delete and reorder pages, read text, render a page to png, audit faux redactions, script its 163 tools. Use for "merge these two PDFs", "extract pages 1 and 3 from this PDF", "rotate page 1 and drop page 2", "what…
craft-data-visual
Turn data into a finished visual with ArtCraft apps - GridCraft workbook with a formula-error read-back, chart, VectorCraft/DesignCraft infographic, DeckCraft slide; verified and resumable. Use for "make a chart from this spreadsheet", "build an infographic from these numbers", "put the revenue chart on a slide".