# DISCLOSURE - pi-web (@pixu1980/pi-web)

## Purpose

pi-web is a web scraping and content extraction tool for the pi coding agent.
It fetches URLs, converts HTML pages into clean, context-ready Markdown, and
returns the result to the LLM for further processing.

## Dual-Use Capabilities

1. **Arbitrary URL fetching**: pi-web accepts a URL from the LLM/user and
   performs an HTTP GET request to fetch the page content. This capability can
   be used to access both public and (misconfigured) internal resources.

2. **Configurable SSRF allowlist**: The built-in SSRF guard blocks private,
   loopback, link-local, and reserved IP ranges by default. However, the
   allowlist is configurable (e.g., to permit local development servers),
   which could be exploited if misconfigured.

## Legitimate Intended Use

- Fetching public documentation, API references, and web pages for the LLM
  to read and reason about.
- Converting web content to Markdown for inclusion in the agent's context.
- Accessing local development servers (e.g., `localhost:3000`) during
  development with explicit user configuration.

## Mitigations

- **SSRF guard enabled by default**: Blocks all private/reserved IP ranges.
- **Per-hop redirect validation**: Each redirect target is re-validated.
- **Configurable timeout and byte cap**: Prevents resource exhaustion.
- **User-controlled allowlist**: Only the user can whitelist IP ranges via
  explicit configuration; no programmatic bypass exists.