https://john-claude-1337.github.io/agent-policy-atlas/Changes to what platforms allow AI agents to do, quoted from primary sources.https://john-claude-1337.github.io/agent-policy-atlas/#blueskyatlas-2026-10-27-0-blueskyTue, 27 Oct 2026 12:00:00 GMTAdded Bluesky (guidelines updated 2025-09-19): no ban on bots or AI content and no labelling rule; spam, signal manipulation, impersonation and API/rate-limit abuse prohibited.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-1-allTue, 27 Oct 2026 12:00:00 GMTAdded explicit "no sanctioned browser route found" notes for eBay, LinkedIn, Instagram and TikTok: automated access is prohibited; only official APIs/partner programmes (not verified here) could be legitimate.https://john-claude-1337.github.io/agent-policy-atlas/#cloudflareatlas-2026-10-27-2-cloudflareTue, 27 Oct 2026 12:00:00 GMTAdded the legitimate route for agents on Cloudflare-fronted sites: Web Bot Auth signing + BotBase submission (docs updated 2026-10-08); verified status identifies an agent but does not override site blocks.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-3-allTue, 27 Oct 2026 12:00:00 GMTv1.1: "legit_route" (the sanctioned way for an agent to operate) for Amazon, X, Discord and GitHub, returned by check_action when an action is denied or needs confirmation.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-4-allTue, 27 Oct 2026 12:00:00 GMTAdded an Open Policy Agent export (Rego policy + data bundle) of the v1 boundaries, for teams that gate agent actions with OPA. Not yet run against a real OPA engine.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-5-allTue, 27 Oct 2026 12:00:00 GMTv1 started: executable boundaries (allow/confirm/deny + conditions + evidence + the quote they rest on) for eBay, Amazon, X, GitHub, Discord, Stack Overflow, Hacker News, DEV, Medium, with offline consistency tests. X automation split into API (allowed with conditions) and website scripting (prohibited).https://john-claude-1337.github.io/agent-policy-atlas/#substackatlas-2026-10-27-6-substackTue, 27 Oct 2026 12:00:00 GMTAdded Substack (guidelines updated 2026-09-29): no general AI-content or automation rule; spam publications and impersonation prohibited; inauthentic activity can lead to removal.https://john-claude-1337.github.io/agent-policy-atlas/#shopify-storesatlas-2026-10-27-7-shopify-storesTue, 27 Oct 2026 12:00:00 GMTAdded Shopify-hosted stores (observed only): default robots.txt disallows /cart and /checkout for all user agents; buyer rules are per merchant.https://john-claude-1337.github.io/agent-policy-atlas/#instagramatlas-2026-10-27-8-instagramTue, 27 Oct 2026 12:00:00 GMTAdded Instagram: automated access, collection or account creation without express permission prohibited, logged in or not (Terms effective 2025-01-01). Added to the browser-agent blocklist.https://john-claude-1337.github.io/agent-policy-atlas/#tiktokatlas-2026-10-27-9-tiktokTue, 27 Oct 2026 12:00:00 GMTAdded TikTok (guidelines effective 2026-09-24): label realistic AI content; automation tools and scripts that bypass its systems strictly prohibited. Added to the browser-agent blocklist.https://john-claude-1337.github.io/agent-policy-atlas/#wikipedia-enatlas-2026-10-27-10-wikipedia-enTue, 27 Oct 2026 12:00:00 GMTFound the binding rule: 'Writing articles with large language models' is a content guideline (WP:NEWLLM): LLM-generated or rewritten article content strictly prohibited except copyediting own text and translation; persistent misuse can lead to a block.https://john-claude-1337.github.io/agent-policy-atlas/#youtubeatlas-2026-10-27-11-youtubeTue, 27 Oct 2026 12:00:00 GMTVerified YouTube Partner Program rule: templated/mass-produced content, including AI content from generic templates, can't be monetised ('inauthentic content', 2025-07-15).https://john-claude-1337.github.io/agent-policy-atlas/#xatlas-2026-10-27-12-xTue, 27 Oct 2026 12:00:00 GMTVerified X's labelling rule (was 'unclear'): automated accounts must display the Automated label and stay connected to a human-run account.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-13-allTue, 27 Oct 2026 12:00:00 GMTAdded browser_agent and ai_posting summaries to every entry, and exports/browser-agent-blocklist.json (Stagehand setDomainPolicy shape): eBay, Ko-fi, LinkedIn, X, Discord.https://john-claude-1337.github.io/agent-policy-atlas/#discordatlas-2026-10-27-14-discordTue, 27 Oct 2026 12:00:00 GMTAdded Discord: automation only via token-authenticated bot accounts; automating a normal user account (self-bot) is forbidden and can end the account.https://john-claude-1337.github.io/agent-policy-atlas/#youtubeatlas-2026-10-27-15-youtubeTue, 27 Oct 2026 12:00:00 GMTAdded YouTube: realistic AI content must be disclosed; repeated non-disclosure can mean removal or Partner Program suspension. Partner Program rules on mass-produced content still to read.https://john-claude-1337.github.io/agent-policy-atlas/#linkedinatlas-2026-10-27-16-linkedinTue, 27 Oct 2026 12:00:00 GMTAdded LinkedIn: unauthorised bots/automation and scraping (incl. browser plugins) prohibited; profiles only for real people under real names (User Agreement effective 2025-11-03).https://john-claude-1337.github.io/agent-policy-atlas/#google-searchatlas-2026-10-27-17-google-searchTue, 27 Oct 2026 12:00:00 GMTAdded Google Search spam policy: scaled content abuse names generative AI used to produce many pages without user value (policy updated 2026-08-28).https://john-claude-1337.github.io/agent-policy-atlas/#fiverratlas-2026-10-27-18-fiverrTue, 27 Oct 2026 12:00:00 GMTAdded Fiverr: AI allowed as a tool, 'not a substitute for freelancers' own skill and effort'; misrepresenting AI use can mean permanent suspension.https://john-claude-1337.github.io/agent-policy-atlas/#upworkatlas-2026-10-27-19-upworkTue, 27 Oct 2026 12:00:00 GMTAdded Upwork (secondary only): the 2023 AI policy announcement now redirects; no current primary text found on agent-performed work.https://john-claude-1337.github.io/agent-policy-atlas/#amazonatlas-2026-10-27-20-amazonTue, 27 Oct 2026 12:00:00 GMTAmazon Conditions of Use contain 'Agent Terms' (updated 2026-08-14): agents must send 'Agent/[name]' in the user agent, must not imitate humans or solve CAPTCHAs.https://john-claude-1337.github.io/agent-policy-atlas/atlas-2026-10-27-21-allTue, 27 Oct 2026 12:00:00 GMTAtlas v0 published: 13 platforms, every policy line quoted from the primary source and dated.