{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiakmtlewwhbcu4mloopdk6k3gzm5lgle5ehqqiygmtbjiacydfqa4",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mox4zmfmxu52"
  },
  "path": "/t/codex-and-coding-agents-need-stronger-scope-control-not-only-better-code-generation/1384420#post_2",
  "publishedAt": "2026-06-23T08:31:27.000Z",
  "site": "https://community.openai.com",
  "textContent": "I agree with the framing. The missing control surface is often not “better generation”, but a small contract the agent can be checked against while it works.\n\nThe pieces I would want to see first-class are:\n\n  * a scope ledger: files/directories the agent is allowed to touch for this task;\n  * an explicit non-goal list: things that may look related but must not be changed;\n  * a diff budget: maximum number of files or behavioral surfaces before the agent must stop and ask;\n  * a stop condition: what evidence counts as done;\n  * negative checks: tests or inspections proving it did not refactor adjacent code, change public API, or alter config unexpectedly.\n\n\n\nThe important part is that these should be machine-checkable, not only written as prose in the prompt. Otherwise the model can “agree” with the boundary and still drift once it starts solving.",
  "title": "Codex and coding agents need stronger scope control, not only better code generation"
}