Document captcha telemetry fields and outcomes

[Sayan-]

Summary

• Document captcha provider, task kind, and the deprecated classification field, including how to correlate observed and inferred challenge results.
• Update the captcha gate example to handle both current and older events, and to distinguish a completed image round from a returned token.
• Document event type filtering for live telemetry streams.

Validation

• Extracted the complete TypeScript example from the four step snippets and confirmed they match.
• Parsed the extracted TypeScript example with Bun and checked the diff for whitespace errors.
• Mintlify preview deployment passed, and the categories page loaded with the new sections.

---

Note

Low Risk
Documentation and example cookbook changes only; no production code or API behavior changes in this diff.

Overview
Expands captcha telemetry docs and aligns the captcha-gate cookbook with the newer event model.

On Categories, the captcha section is restructured around captcha_provider, task_kind, and deprecated captcha_type, with tables for payload fields, per-provider emissions (observed vs inferred captcha_challenge_result), correlation rules (task_id vs challenge_id), and updated outcome meanings including inferred. The sensitivity table now mentions provider and task kind.

Pausing for Captcha Solves updates captcha-gate.ts to parse those fields (with fallbacks for older events), key joinable IDs as task: / challenge:, accept inferred challenge results on task_id, wait for challenge results via expectsChallengeResult, and surface distinct agent messages for token tasks, image rounds, observed clears, and inferred token results.

Streaming documents server-side filtering by passing type when opening the stream (and the same filter on the read-events HTTP API).

^{Reviewed by Cursor Bugbot for commit d9f0738. Bugbot is set up for automated code reviews on this repo. Configure here.}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
Kernel	🟢 Ready	View Preview	Oct 14, 2026, 6:50 PM

💡 Tip: Enable Automations to automatically generate PRs for you.

[akxue]

some legibility nits

this is fairly tough to read imo, and maybe more suited as small tables?

the captcha section itself has a ton of info packed into it, what do you think about breaking it out into multiple sections?
my suggestion would be like

• captcha event fields
• what each captcha provider emits: the emission table and observed vs inferred results
• correlate captcha tasks and challenges: joining on task_id and challenge_id, why relay tasks have no challenge_id, etc
• captcha challenge outcomes: statuses per provider and the outcome table

some minor accuracy nits my guy helped me catch as well

• browsers/telemetry/categories.mdx:90 only relay image rounds report other
• browsers/telemetry/categories.mdx:96 every row in the challenge_id column now says "can be absent", and the sentence explaining why was removed (line 126). human always has it (press-hold-worker.ts), hcaptcha has it only on extension image rounds, and relay tasks never do (relay.go:360). that's why an observed recaptcha_v2 result can't join its tasks, which the gate depends on. i think we should keep the yes / sometimes / no split and say why relay tasks lack it

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit d9f0738. Configure here.}

[Sayan-]

thanks for the feedback! ready for another round @akxue

[akxue]

ty! this is looking nice