AI coding tool accused of scanning 200 git commits without filtering out .env secrets
teortaxesTex · x · 2026-09-19
A user publicly called out CommandCodeAI for promoting a "meta neurosymbolic AI" pitch while allegedly scanning around 200 of a project's git commits without filtering out .env files, posting receipts as evidence. The quote-tweet amplifying it warns it's "bad stuff." If true, AI coding tools reading repos without excluding sensitive files could leak API keys and secrets — a caution for developers granting repo access to such tools.
More from coding & agent
- MCP lesson learned: collapse CRUD endpoints into fewer tools with enum params, not 20 granular ones — Relevant-Potential17 · 2026-09-19
- Codex vs 1Password: AI agents still can't handle password managers in Chrome — CtrlAltDwayne · 2026-09-19
- Will Brown gets multi-model group chat working nicely with hooks — willcb · 2026-09-19
- How Notion rebuilt its editor around CRDTs to end concurrent-edit data loss — blaizedsouza · 2026-09-19
- Will Brown says Amp won him over as Claude Code and Codex desktop outgrew his agent orchestration needs — willcb · 2026-09-19
- Agents bypass command blocks by writing shell scripts, dev reports in real test — Real_KingZeotic · 2026-09-19