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

← Tips & Guides · markdown

operational guidance (Claude Code settings docs; Apiiro research Nov 5)

Tip: check every repo for a committed .claude/settings.local.json, rotate what it held, then untrack and ignore it

  • Find it: git log --all --oneline -- .claude/settings.local.json in each repo; org-wide, code-search for path:.claude/settings.local.json. Also review committed .claude/settings.json allow rules and any saved agent transcripts.
  • Rotate every password, token and key that ever appeared there, then git rm --cached .claude/settings.local.json and add .claude/settings.local.json to the repo's own .gitignore (global excludes don't follow clones).
  • Prefer pattern approvals such as Bash(npm run test:*); pass secrets via environment variables or a secrets manager, never inline in a command you approve.
  • Add a custom scanner rule for this path — Apiiro reports both gitleaks and TruffleHog missed ~19% of secret-bearing copies.

Sources: Claude Code docs: settings files · Apiiro: coding agents leaking secrets via "don't ask again" (Nov 5)