SI Tools 2026 10 23 Your arguments ran as shell

SI Tools 2026 10 23 Your arguments ran as shell You paste an issue title into GSD and the apostrophe ends the quoting. GSD's own PR says typed arguments "ended the quoting and ran as shell" in 133 places across 58 workflow files. That sets this week's theme: data that got executed. Open GSD Core 1.18.0 fixes that injection. It also blocks a git -c alias trick that printed .env past the secret guard, stops a Brave API token from following redirects, and makes ambiguous artifact checks fail closed. Claude Code 2.1.302 through 2.1.309 (eight releases) stops the sandbox from running your shell startup files outside the sandbox before every command, and stops sandboxed commands from editing --mcp-config. On Windows, PowerShell guard hooks that exit 2 actually block now, and exFAT/FAT32 projects can be written again. Quiet default changes: allowedMcpServers now loads some servers it used to filter, headless sessions start in default mode in some setups, and /rewind no longer undoes sed -i. The Claude API's first new version since 2023 brings C# SDK 13.0.0, Apache 2.0 licensing, and 400s for temperature. Buildomator 5.0 finally ships, and the cliff is deferred. Superpowers 7.0.1 retunes brainstorming and warns about a 20,000-character compaction limit. Grok Build 1.0.62 stops reporting success when it gave up. Grok Bot 0.75.2 stops letting typing answer a question card. X segment: Agent Skills goes to the Agentic AI Foundation, Claude apps go auto-mode-by-default next week, Claude Duet launches, and Grok Bot picks up phone apps and 1Password. Every item is labeled. X posts referenced: Confirmed: @ClaudeDevs Agent Skills donated to the Agentic AI Foundation https://x.com/ClaudeDevs/status/2113328526... Confirmed: @bcherny Agent Skills donation https://x.com/bcherny/status/2113333859907... Partly confirmed (announced, not live): @claudeai auto mode default in Claude apps https://x.com/claudeai/status/211336084540... Confirmed: @ClaudeDevs API version 2026-10-21, first since 2023 https://x.com/ClaudeDevs/status/2113013247... Confirmed: @claudeai Claude Duet research preview https://x.com/claudeai/status/211298466425... Confirmed: @_catwu Claude Duet https://x.com/_catwu/status/21129887515763... Confirmed: @claudeai Claude Security continuous scans https://x.com/claudeai/status/211296224821... Confirmed (re-announcement): @ClaudeDevs OSS Scanner https://x.com/ClaudeDevs/status/2112299602... Confirmed (self-reported): @ClaudeDevs ~90% of Claude Code written by Claude https://x.com/ClaudeDevs/status/2111168901... Confirmed: @bot reports, memos, dashboards (0.75.2 Pages) https://x.com/bot/status/2112992872857252082 Partly confirmed: @bot phone Calendar, Reminders, Contacts https://x.com/bot/status/2113393684209176718 Confirmed (rolling out): @bot 1Password integration https://x.com/bot/status/2113332983226831339 Confirmed (self-reported): @bot 204 tickets in about 10 minutes https://x.com/bot/status/2112645531076096330 Partly confirmed: @mattyp What's New in Grok Bot chapters https://x.com/mattyp/status/21125501588859... Claim only: @elonmusk RT fastest product growth https://x.com/elonmusk/status/211285090116... Unverified: @zxiiro Grok Build 402 at 33 percent meter https://x.com/zxiiro/status/21121975870087... Unverified: @dlqnt_xyz Grok Build 402 at 50 percent meter https://x.com/dlqnt_xyz/status/21122234825... Opinion (context for the 7.0.1 brainstorming change): @kaspars_z Superpowers self-invokes aggressively https://x.com/kaspars_z/status/21124970570... X posts are treated as leads only. Each was checked against primary sources (changelogs, docs, release notes) and labeled. AI-generated audio deep dive (NotebookLM).