Only the latest commit on main and the most recently tagged release get security fixes. This project is pre-1.0 (see CHANGELOG.md) — older tags are not backported.

Please report security issues privately — open a GitHub security advisory on this repo, or email support@2captcha.com — rather than a public issue. Include the version/commit and a minimal reproduction.

We aim to acknowledge a new report within 5 business days, confirm whether it's in scope, and share a fix timeline once it's confirmed. If a report turns out to be a real, exploitable issue, we'll credit the reporter in CHANGELOG.md unless they'd rather stay anonymous.

In scope: this repo's own code and dependencies — credential handling, injection risks, anything that would make this tool leak a secret, execute untrusted code, or misreport what it actually did (a complete status on a run that silently dropped data would count, for example).

Out of scope: vulnerabilities in perplexity.ai itself — report those to Perplexity directly, not here. This tool only ever reads an article's own already-published content; it never touches Perplexity's live "Ask AI" feature, never logs in, and never submits anything — a report that this tool "can't" do any of those is a scope statement, not a vulnerability. Similarly, a robots.txt-disallowed path being skipped (never requested at all — see page_parser.is_disallowed_path) is documented, intended behavior, not a bug to report here.

• TWOCAPTCHA_KEY, PERPLEXITY_PROXY, PERPLEXITY_CDP_ENDPOINT belong in .env or the environment (see env_config.py). The matching --twocaptcha-key / --proxy / --cdp-endpoint flags exist, but a credential on a command line ends up in shell history and process lists, so the docs never use them. Credentials are never logged in full: every log line masks them, keeping only host:port.
• This project's own code doesn't phone home. The only network calls it makes are to perplexity.ai and, when configured, api.2captcha.com / cb.2captcha.com. One caveat on the Selenium engine: Selenium's own bundled Selenium Manager sends anonymous usage stats to plausible.io by default whenever it launches a browser, which selenium_scraper.py disables on your behalf (SE_AVOID_STATS=true, set as a default rather than forced, so it never overrides a value you set yourself) so this project's "nothing phones home" claim actually holds. Selenium Manager can still make a separate network call to resolve a matching chromedriver version if one isn't already reachable on PATH / SELENIUM_CHROME_BIN — set that variable to a Chrome/Chromium binary you already have to avoid it entirely (see selenium_scraper.py's module docstring).
• This is a scraper, not an account tool: it never logs in, never touches an authenticated perplexity.ai session, and never writes anything back to perplexity.ai — no comments, no follow-up questions submitted through the live "Ask AI" feature, ever.