pwn.watch Tracks Live Security Incidents

// 1h agoPRODUCT LAUNCH

pwn.watch Tracks Live Security Incidents

pwn.watch is a live security-incident dashboard covering CVEs, supply-chain compromises, patch status, fixed versions, and advisories. Its AI/LLM tooling feed makes developer risks like the LiteLLM compromise easier to monitor and automate.

// ANALYSIS

This is practical security infrastructure for developers: focused, searchable, and built for turning incident data into action.

–Tracks affected versions, fixes, severity, and patch status instead of merely listing CVEs
–JSON, RSS, llms.txt, and per-tag feeds make the service unusually friendly to coding agents and automation
–Its AI/LLM category covers real developer-facing threats, including compromised packages and vulnerable MCP tooling
–The main risk is coverage and editorial trust: a curated feed is only as useful as its source verification and update speed

// TAGS

pwn-watchsecurityobservabilitysearchapiautomation

DISCOVERED

1h ago

2026-10-22

PUBLISHED

4h ago

2026-10-22

RELEVANCE

7/ 10

AUTHOR

mrkd

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

OPEN SOURCE20m ago

Agent Skills Moves to Neutral Stewardship

Anthropic is donating Agent Skills to the Linux Foundation-backed Agentic AI Foundation, the same neutral home now stewarding MCP. The open SKILL.md standard has already reached more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot, and VS Code.

UPDATE1h ago

Macroscope Approvability Dashboard Makes AI Approvals Measurable

Macroscope’s new dashboard shows which pull requests its Approvability agent approves, escalates, or skips, helping teams understand approval behavior and measure how policy changes affect the results. Teams can tune rules for protected paths, routine changes, correctness thresholds, and required checks.

MODEL1h ago

Gemma-4 Giftige 31B v1 Ships GGUF

0xSojalSec has published GGUF quantizations of the 31B Gemma-4 Giftige-Blume merge, targeting uncensored roleplay and creative writing. Files range from 12GB Q2_K to 32.7GB Q8_0, making local inference accessible across a range of hardware.