kits: the prompt states the output contract, so a weak model writes the file instead of answering in chat - #32
Merged
Conversation
…he file instead of answering in chat A Sheets run on codex with gpt-5.4-mini read the sheet-design Skill, researched the question and answered with a table in the conversation; ./sheet.json was never written and the person saw "No sheet has been written for this conversation yet." The Skill says the file is the deliverable, but the prompt let a research question read as a chat task. The sheets, slides and dashboard prompts now say what the person sees (only the file), what every request ends with (that file written or updated), and that an answer living only in chat is invisible to them. Sheets adds the shape a research question takes. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A Sheets run on codex with gpt-5.4-mini read the sheet-design Skill, researched the question and answered with a table in the conversation;
./sheet.jsonwas never written and the person saw "No sheet has been written for this conversation yet." The Skill says the file is the deliverable, but the prompt let a research question read as a chat task.The sheets, slides and dashboard prompts now say what the person sees (only the file), what every request ends with (that file written or updated), and that an answer living only in chat is invisible to them. Sheets adds the shape a research question takes: rows for the things found, columns for the facts, an agent column for the work to repeat per row.
Verified on the hosted service with the exact prompt that failed (codex, gpt-5.4-mini): see the follow-up comment.
🤖 Generated with Claude Code