Repository navigation
A draft pull request opened with the description of the state script, and nothing of the session's - #95
Merged
Conversation
… and nothing of the session's In 3 of the 6 runs of the fourth campaign of the evaluations that had a `gh` to call, the draft pull request was opened with the text `surface-status pr-body` prints followed by a line of attribution, where the prompt says "and nothing else" (#93). The stand-in `gh` of #91 let it be seen for the first time. The two commands did not ask for the same thing. `/surface-plan` named `gh pr create --draft` "with [...] that description" and `gh pr edit --body`, and left the session to carry the text: every session wrote it to a file of its own first, and three added a footer to it. `/surface-execute` gives the output of the script to `gh pr edit --body-file -` on its standard input, and no refresh of the three runs played whole carries one. `/surface-plan` now does the same: the output of the script is given to `gh` on its standard input, with `gh pr create --draft --body-file -` when the branch has no pull request and `gh pr edit --body-file -` when it has one. Measured on `05-fine-cap`, where the footer was written in two runs of two: three runs played to the hand over, three drafts opened from the standard input with the description alone, and at every hand over the description is the one the script prints. 19 sessions, 4.94 USD.
PierreMardon
force-pushed
the
fix/pull-request-description-piped
branch
from
October 5, 2026 11:17
57b4e3a to
35d9d1e
Compare
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What changes
/surface-plangives the output ofsurface-status pr-bodytoghon its standard input, as/surface-executealready does at every refresh:gh pr create --draft --body-file -when the branch has no pull request,gh pr edit --body-file -when it has one.Why
In 3 of the 6 runs of the fourth campaign of the evaluations that had a
ghto call, the draft pull request was opened with the text the state script prints followed by a line of attribution, where the prompt says "and nothing else" (#93). The stand-inghof #91 let it be seen for the first time.The two commands did not ask for the same thing:
/surface-plannamedgh pr create --draft"with [...] that description" andgh pr edit --body, and left the session to carry the text. Every session wrote it to a file of its own first, and three added a footer to it./surface-executegives the output of the script togh pr edit --body-file -on its standard input, and no refresh of the three runs played whole carries a footer.A sentence did not turn that default of the session off every time. A command that leaves it no text to write does.
Files
skills/surface-plan/SKILL.md, "Committing, pushing, the pull request": the two calls, and their reason.tests/prompts/test_surface_plan.py: the contract test of the draft holds the two calls, the reason, and thatgh pr edit --bodyis gone.Measured
On
05-fine-cap, where the footer was written in two runs of two, three runs played to the hand over from this branch (.evals/fix-93/, out of git; 19 sessions, 4.94 USD):pr_describedat the hand overgh pr create --draft --title "..." --body-file /tmp/prbody.mdgh pr create --draft --title "..." --body-file -In the run that took a correction, the description of the second revision is set by
gh pr edit --body-file -, with no footer either.Not measured: the other cases, and a real
gh, whose--body-file -the stand-in imitates.Refs #93