{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreid7vhwn7wayiqvjqgnippppscsrjwkow6uojsafqx2nclux5533pe",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mnzfli4s2bk2"
  },
  "path": "/t/if-unsure-ask-never-guess-ai-agent-pre-execution-checklist/176632#post_7",
  "publishedAt": "2026-06-11T13:35:41.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "AI is basically in a waiting state.\n\nIt does not know when the user will ask for something, what the user will ask for, or what meaning the user will attach to that request.\n\nWhen a user instruction arrives, the AI first tries to interpret it as an executable Action.\n\n**This is where the most dangerous inference occurs.**\n\nNot in the Provider Checklist first, but in **C1 and C2.**\n\n**C1 — When, or in what case, does the user want this Action to run?**\n**C2 — What does the user mean by this Action?**\nC3 — Which Provider Action will actually be called?\n\nAmong these, the AI must be most careful not to infer C1 and C2.\n\nBecause **C1 and C2 are directly connected to user intent**.\n\nIf the AI guesses here, it may turn something the user did not ask for into something that appears executable.\n\nThis is not merely a safety question.\nIt is first a question of **intent completeness.**\n\nBefore asking whether an Action is safe, the system must know what Action the user actually intended.\n\nThe Provider Checklist has a different role.\nIt does not define the user’s intent.\n\nIt defines what must be confirmed for this specific Action to be executed accurately and safely: boundaries, required conditions, completion criteria, stop conditions, and execution limits.\n\nIn other words:\n\n**The Provider Checklist says what must be checked before this Action can be executed.**\n\nThe User Checklist has two roles.\n\nFirst, it completes **C1 and C2** by clarifying the user’s intent for this specific case.\nSecond, it supplies the runtime context required by the Provider Checklist.\n\nSo the User Checklist can also improve accuracy and safety.\nBut its first and most important role is to make the **user’s intent explicit**.\n\nThe structure is therefore not:\nUser says something → AI guesses the Action → Provider Checklist checks safety.\n\nThe structure should be:\nUser says something → AI identifies missing intent → User completes C1 and C2 → Provider Checklist confirms execution requirements → AI executes only when the instruction is complete.\n\nThe first unknown is not safety.\n\nThe first unknown is intent.",
  "title": "If unsure, ask. Never guess. — AI Agent Pre-Execution Checklist"
}