Stroq 0.30.0 locally audits coding-agent sessions for dictated commands and leaked keys
Open-source Stroq shipped 0.30.0 on November 5. Running npx @stroq/cli reads the last 30 days of Claude Code and Codex sessions on the machine, replays them through its policy in a throwaway home (nothing executes, nothing is sent), and prints a card: how many commands the agent ran, how many of those came from text it read (a README, page, issue, or tool result), which key names left the machine, and which repositories run code as soon as an agent opens them. Moments worth a look sit under the counts; stroq show N prints the traced call with secrets replaced by names. The release also adds rule STROQ-2026-00014, which flags priority-style tasks slipped into MCP results, pages, or repo files ahead of the user's own request. It is Apache 2.0 and sends no telemetry by default.
Why it matters
Coding agents increasingly act on text they read. This release turns a local session audit into a one-command card and adds detection for tasks slipped into tool or file content.
Sources1
Written by dalao.ai: the title, summary and why it matters are our own words. The original reports are the record; their publishers keep authorship.