Write prohibitions that actually bind
Starts from failures that happened, not ones you imagine.
| Category | Using AI › Writing prompts |
|---|---|
| Tags | DraftingAnalyzingReviewing |
Write the "do not" rules for this prompt. ***The ones that actually bind.*** **Premise: most prohibitions in prompts do nothing, and the ones that matter are missing.** 1. **What has actually gone wrong** in my outputs so far. ***Start from failures, not from imagined ones*** 2. **Per failure, write the rule that would have prevented it** — ***naming the specific move, not the category*** - ✗ "Do not be verbose" — ***not checkable*** - ✓ "Do not restate my request before answering" — ***a specific move*** 3. **Mark which rules will bind and which are decoration:** - ***Rules about what to do when information is missing*** — **these bind** - ***Rules forbidding a specific, nameable action*** — **these bind** - Rules about tone or quality — ***usually decoration*** 4. **Where a prohibition needs an alternative** — ***"do not X" leaves a gap; "do not X, do Y instead" closes it*** **Then:** - **The rules, ordered by how often they will actually fire** - ***Which ones to cut*** — **a long list of prohibitions dilutes the ones that matter** - **Where two rules conflict under some input**
After pasting, fill in the fields at the bottom (My prompt · What has gone wrong)
What this prompt does
Most prohibitions in prompts do nothing because they are not checkable. The ones that bind name a specific move or say what to do when information is missing — this derives them from real failures.
More in this category
| 4 Optimized Versions of A Prompt (in Arabic) | |
| $500/Hour AI Consultant Prompt | |
| Adaptive Thinking Framework | |
| Agent Organization Expert | |
| AI Kickstart prompt |