Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions docs/plan.md
Original file line number Diff line number Diff line change
Expand Up @@ -104,6 +104,8 @@ Detect silly prompts, and maybe show a banner that it's garbage input.
Currently PlanExe fails to recognize the comedic/weird prompts. Prompt has a playful tone, that isn't your typical enterprise project.
It yields a non-sense plan. Instead it would make more sense to respond with a similar playful tone.

**Self-contradictory prompts:** A prompt can ban its own subject. The observed case: a plan explicitly about *"AI agents"* that listed `AI` among its banned words. The constraint checker then (correctly) rejects every generated lever as violating "Do not use AI", `IdentifyPotentialLevers` produces zero levers, and the run dies several stages later with a misleading `No input levers to deduplicate`. As of PR #766 this fails loud at the source and names the offending constraint, but the real fix is to catch the contradiction up front — at constraint extraction or the redline gate — and tell the user before any levers are generated, so no tokens are spent on a run that cannot succeed. The detector should be language-agnostic and not rely on exact word matches (a banned concept can reappear as a synonym), and it should distinguish a genuine contradiction from a narrow, legitimate exclusion.

**Ask for expert help:** Establish contact between people, for reviewing a plan, for executing the plan, for getting funding.
The “Ask for expert help” section, serve the content from planexe.org. Either as an iframe or as javascript or be generated dynamic?

Expand Down
Loading