- All ~400 dorks and queries across Google, GitHub, Shodan, and Censys now use a
KEYWORDplaceholder (or a native scope filter:org:,site:,hostname:,net:,autonomous_system.organization) in place of hard-coded secret-seeking value strings. - Service fingerprints (product titles, API paths, ports, env-var names, key prefixes) are intentionally kept β they identify the service/format, not a victim's secret.
- README opens with a new "
KEYWORDConvention β Read This First" section. - CONTRIBUTING makes the
KEYWORDconvention mandatory for new dorks.
- Grok
site:grok.com/shareβ re-confirmed π’ still broadly indexed. - ChatGPT
site:chatgpt.com/shareβ kept π‘ degraded; de-indexing ongoing since the Aug 2025 feature removal, but cached/third-party-archived copies persist and spot checks still return occasional live results. Bing/DuckDuckGo may retain more than Google. - HuggingFace Spaces β re-confirmed π’ active.
- GitHub syntax β re-confirmed web Code Search uses
path:; legacyfilename:survives only via REST API v3. - ChromaDB β re-confirmed
/api/v2, port 8000, no auth by default, binds localhost. - Default ports & key prefixes β re-verified against official docs.
- Content scope (CVEs, threat-intel, tools, Sigma rules) carried forward unchanged from v1.2.0; this release is a parameterization + verification refresh, not a content expansion.
- A
KEYWORD_BANK.mdof the previous secret-seeking terms was preserved out-of-tree for upcoming tooling work (not published in the repo).
- MCP systemic RCE (Ox Security, 10+ CVEs), Claude Code source-map leak, ChatGPT DNS exfiltration, GitHub Copilot wormable RCE (CVE-2025-53773), nginx-ui MCPwn (CVE-2026-33032), DeepSeek ClickHouse exposure, 300K+ ChatGPT creds, additional exposed Ollama instances, Mythos/Glasswing.
- Initial public release.