tag: Code-Execution · 7 items
- Engineer — Plan: Developers using OpenCode on Windows or WSL should update to version 1.18.33; the vulnerability allows a compromised or rogue MCP server to execute code under the developer’s credentials at sign-in. No KEV listing or public PoC reported, so patching on normal schedule is appropriate rather than emergency response.
- SOC/IR — Learn: This introduces a novel attack surface where AI coding agent tool integrations (MCP servers) can be used as a code execution vector against developer workstations; no IOCs, TTPs, or active exploitation reported yet, so no immediate detection work is warranted.
- Leader — Skip
- Engineer — Plan: If your environment runs LibreOffice or OpenOffice with Java support enabled, audit those deployments and disable Java as a mitigation now; monitor for a vendor patch and apply it promptly when released. No active exploitation yet, but the attack surface (malicious file opened by a user) is realistic.
- SOC/IR — Learn: No active exploitation or published IOCs yet, so there is nothing actionable to hunt; worth tracking in case threat actors weaponize this for phishing lure campaigns delivering malicious spreadsheets.
- Leader — Skip
- Engineer — Learn: Topics like AI coding tools leaking repos and one-click code execution are directly relevant to engineering environments, but this digest provides no CVEs, affected versions, or concrete actions — review the linked full roundup to identify any items touching your stack.
- SOC/IR — Learn: AI search poisoning and trusted-path abuse are worth tracking as emerging lure and delivery techniques, but the summary contains no IOCs, TTPs, or detectable behaviors — check the full article for any specific campaign details.
- Leader — Skip
- Engineer — Plan: A public PoC exists but EPSS is near zero and it is not KEV-listed; exploitation requires REPLICATION attribute (a non-default privilege), so immediate fire-drill pressure is low. Patch PostgreSQL to 18.6, 17.11, 16.15, 15.19, or 14.24, and audit which accounts hold the REPLICATION attribute in your instances.
- SOC/IR — Learn: No active exploitation is reported and no IOCs are available; the public PoC is useful context for understanding how a compromised REPLICATION account could achieve OS-level execution, informing future detection logic around abnormal logical decoding activity, but no immediate hunt is warranted.
- Leader — Skip
- Signals: CVE-2026-6471 — CISA KEV: not listed, EPSS 0.00, public PoC on GitHub
- Engineer — Plan: Four of the seven affected agents remain unpatched, making this an active exposure for any team whose developers clone untrusted repos while running AI coding assistants. Audit which agents (Claude Code, Codex CLI, Cursor, etc.) are in use, update those that have received patches, and enforce policy against running agents against repositories from untrusted sources until remaining fixes ship.
- SOC/IR — Learn: This research introduces a new attack class—git-config-triggered code execution via AI agent trust boundaries—that is worth understanding for future detection work on developer endpoints, but no IOCs, exploited campaigns, or mappable TTPs are published yet to act on immediately.
- Leader — Plan: With four tools still unpatched, any organization where developers use CLI AI coding agents carries uncontrolled supply-chain risk from malicious repository clones. This quarter, inventory which agents are deployed, confirm patched versions are standardized, and establish a policy on approved repositories before AI agent use.
- Engineer — Act: Public PoC exists for a heap overflow triggered by opening a crafted XZ archive in 7-Zip, a tool common in dev workstations and CI/CD pipelines; patch all 7-Zip installations to 26.02 and audit any automated pipeline steps that extract XZ archives unattended.
- SOC/IR — Plan: No confirmed in-the-wild exploitation yet, but the public PoC raises urgency; build a detection for anomalous child processes spawned from 7-Zip binaries (7z.exe, 7zG.exe) during extraction, prioritizing CI/CD runners and build servers where archives are processed automatically.
- Leader — Skip
- Signals: CVE-2026-14266 — CISA KEV: not listed, EPSS n/a, public PoC on GitHub
- Engineer — Act: Any developer who opens an untrusted repo in Cursor on Windows is at risk of credential theft (SSH keys, cloud tokens) with no user interaction required — the attack path is fully described, making it practically exploitable now. Update Cursor to the patched version immediately; until confirmed patched, audit recently cloned project directories for unexpected git.exe files and avoid opening untrusted repos in Cursor on Windows.
- SOC/IR — Plan: No active campaign IOCs are reported, but the technique is clear: build a detection for Cursor (or any IDE process) spawning child processes from non-standard project root paths, specifically hunting git.exe executions outside of installed VCS tool directories on Windows endpoints.
- Leader — Plan: Cursor is widely adopted among developer teams; this flaw enables silent credential and source-code compromise via a simple repo-clone workflow. Circulate a developer advisory this week, confirm vendor patch availability, and consider a temporary policy restricting Cursor on Windows for repos from untrusted sources until remediated.