Claude Code 2.1.326 adds an 'execute with care' system prompt section governing destructive agent actions
ClaudeCodeLog · x · 2026-11-18
Claude Code CLI 2.1.326 (367 CLI changes) includes a notable system prompt update: a new "Executing actions with care" section. It instructs the agent to weigh reversibility and blast radius, defaulting to user confirmation before destructive or shared-impact actions (deletes, force-push, git reset --hard, CI/CD changes, sending messages, infra/permission edits). It adds git safety rules (run git status before discard-prone ops, avoid destructive shortcuts like --no-verify, prefer resolving conflicts over discarding, inspect staged files for secrets) and a scope principle: one-time approval never extends beyond the requested scope unless durably authorized in CLAUDE.md. A useful reference template for anyone designing agent safety guardrails.
More from coding & agent
- Code Mode turns 1: every major agent harness has adopted it — and kept the name — ritakozlov · 2026-11-18
- GitHub and Microsoft's MDASH pilot turns thousands of AI vuln findings into 99 verified fixes — mustafasuleyman · 2026-11-18
- OpenAI updates API dashboard with per-model usage, rate-limit and error tracking — OpenAIDevs · 2026-11-18
- 'Proof-Carrying Calls' Idea: Make Agents Certify They Don't Break Your Invariants — tianyin_xu · 2026-11-18
- Devin Platform Adds a 'Poke' Bot Template: An Engineering Agent with Personality — marvinvonhagen · 2026-11-18
- Carbon TTFT now 1.5x faster on Antigravity, Argon to follow — cgarciae88 · 2026-11-18