[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← Threat Hunting · markdown

confirmed research IOCs (Snyk Agent Skill reassessment, Nov 6); critical skills named with install/download counts from skills.sh / ClawHub snapshots

Hunt: malicious agent skills — AIOZ lookalike api-w3stream.attoaioz.cyou, silent ~/.claude/CLAUDE.md plant, slidew / tussahre.pro lures

  • Network: alert on agent/CI egress to api-w3stream.attoaioz.cyou, tussahre.pro, and unexpected skill-hosted validation endpoints.
  • Endpoint: unexpected writes to ~/.claude/CLAUDE.md / ~/.claude/ right after a skill install or 'validate' script; UPX-packed 'fast-validate-*' binaries under skill script dirs.
  • Package: block/flag npx slidew when users expected Slidev; review recently installed ClawHub / skills.sh packages against Snyk's critical list.
  • Full citations and code excerpts are in the Snyk appendix — treat the list as a starting IOC pack, not exhaustive.

Sources: Snyk Research: Re-Assessing the Security of the Agent Skill Eco System (Nov 6)