Agent Prompts Have No Guarantees: Injected Text Can Override Your Instructions
mmitchell_ai · x · 2026-10-23
mmitchellai warns that agents process web pages, files, emails and tool outputs, and text in that data can act as commands the agent follows over yours (prompt injection), so even perfectly worded instructions carry no guarantee of being obeyed.
The same instruction isn't even stable across runs: minor rephrasing changes behavior, and a prompt that works on one model can fail on another, since output is sampled stochastically rather than deterministically.
More from coding & agent
- MCP Tip: One-Line Instructions Beat Long Tool Descriptions, 18/20 vs 3/10 in Evals — colinmcnamara · 2026-10-23
- DynaLM: Open-Source CPU-First LLM Inference Engine Seeks Feedback — Unlikely_Safety_7456 · 2026-10-23
- Software factories as code: Warp runs agents, runners and model routers from markdown/YAML repos — vikvang1 · 2026-10-23
- Dev uses Codex to remotely root a Samsung TV and strip eMudhra root certificates — charliermarsh · 2026-10-23
- Raindrop launches Change Reports to track how agent changes perform in production — HamelHusain · 2026-10-23
- Stop keeping agent sessions local: shared traces enable self-improving agents — vikvang1 · 2026-10-23