Claude

SlopGate

For teams merging Copilot/Cursor/Claude-Code PRs, scans every diff for AI-slop patterns (duplicated boilerplate across files, phantom unused abstractions, copy-pasted error handling) and posts a numeric slop score plus file:line flags before merge.

TrainToggleWatch

For freelancers and consultants who process client data in personal ChatGPT accounts under NDA, checks the 'Improve the model for everyone' toggle daily, auto-flips it back off if OpenAI silently re-enabled it, and issues a weekly PDF opt-out proof log.

StandbySnitch

Runs on a home Raspberry Pi/OpenWrt router and produces a weekly signed log of every connection a smart TV makes to known ACR/telemetry endpoints while the TV is in standby, plus a pre-filled GDPR/CCPA/FTC complaint letter citing the exact timestamps.

HubMirror

For small AI teams whose production code pulls model weights straight from Hugging Face Hub, mirrors every referenced repo+revision to their own S3/R2 bucket within 24h and alerts weekly if the upstream copy is deleted, gated, or silently changed.

PlayBanWatch

Android indie/FOSS devs scan their Play Store listing + APK against a live database of real enforcement triggers (donation links, F-Droid mentions, sideload wording) and get flagged lines cited to the actual precedent case, before Google's next policy sweep suspends them.

ACLSentry

For teams running Tailscale in production, diffs every tailnet ACL policy change within 5 minutes and flags newly-added wildcard destination/port grants before they ship, plus a weekly least-privilege report comparing granted access to actual connections.

CheckpointBench

For teams self-hosting open-weight LLMs (Qwen/Llama/DeepSeek family), auto-runs your own 20-30 prompt eval set against every new checkpoint release and emails a ranked win/loss table within 24h of release so you know whether to switch before burning a day on manual benchmarking.

StegoWipe

Journalists and whistleblowers drop a screenshot or edited image and get a report of hidden device-identifying watermarks (MS Paint/Windows Photos GUID, C2PA manifests) plus a stripped, safe-to-publish copy in under 30 seconds.

MCPScope

For teams connecting third-party MCP servers to Claude Desktop/Code or Cursor, runs the server in a network-monitored sandbox and produces a report of every outbound host and tool call it actually made versus what its manifest declared.

TeenChatDigest

Parents of teens using ChatGPT/Claude/Character.AI get a weekly email digest flagging self-harm, grooming, or cheating-request patterns in their teen's chat exports, with the exact flagged excerpts and a conversation-starter script.

GlassSniff

Detects Meta Ray-Ban/Ray-Ban Display style camera-glasses BLE signatures within ~10m and alerts gym/spa/school staff within 5 seconds so they can enforce no-recording policies in locker rooms and classrooms.

WorkflowFence

For repos that allow GitHub Copilot Autofix, Dependabot, or Snyk auto-PRs, blocks merge of any bot-authored PR that touches CI/CD workflow files or expands permissions/secrets access, and posts the exact lines that triggered the hold.

InjectionWitness

For site owners who moved DNS/CDN providers, nightly-diffs their live HTTP response against an approved script/cookie allowlist and alerts within 24h the first time a host silently injects a new tracker or cookie.

MatchDispute

Job applicants who get a background-check hit that actually belongs to someone with a similar name upload the report and get an FCRA-cited dispute letter, addressed to the specific screening vendor, generated in under 10 minutes.

ModelDriftWatch

Runs the same 50 canary prompts against your production LLM model+version daily and Slack-alerts you within 4 hours if a silent provider update measurably changes output quality, before your users notice.

SkillDrift

For teams running third-party Claude/agent SKILL.md packages in production, watches each skill's source repo and sends a diff alert within 24h of any update flagging new network calls, filesystem paths, or instruction-override phrasing before the next agent run picks it up.

WALGuard

For indie teams running SQLite in production (via Litestream, Turso, fly.io volumes), a lightweight agent runs integrity checks and WAL-health probes every 15 minutes and pages you before silent corruption reaches your backups.

ConsentDial

Gives French telemarketing agencies a pre-dial API that returns ALLOW/BLOCK per lead based on logged consent proof, plus an audit PDF ready for DGCCRF inspection, so a cold-calling team can prove compliance with France's new opt-in-only telemarketing ban in under 200ms per check.

ShareLeakAudit

For ops/IT admins running tl;dv, Fireflies, or Grain across their team, scans every recording's sharing setting weekly and flags any set to 'anyone with the link' before it leaks client or HR data.

RejectionRadar

Indie iOS devs upload their build + App Store Connect metadata and get a report flagging the exact guideline clauses their submission is likely to violate, before they hit Submit.

OralDefenseKit

Turns a student's submitted paper into a printable set of 8-12 paper-specific oral defense questions plus a scoring rubric in under 60 seconds, so instructors can run Denmark-style oral checks without writing questions by hand.

CrawlTollGate

A drop-in reverse-proxy for self-hosted sites that fingerprints AI-training crawlers by UA+ASN, blocks them by default, and emails a weekly report of blocked requests and bandwidth saved.

ScamForward

Forward any suspicious text or email to one address and get a plain-language risk verdict in under a minute, with an optional alert to a family guardian.