review: point read_prompt at the prompt directory, not a file - #484
Open
kees wants to merge 1 commit into
Open
Conversation
The read_prompt tool resolves the name it is given against a base path, so that base has to be the prompt directory: ToolBox is constructed with one everywhere else, and its test reads "technical-patterns.md" relative to it. Worker-run reviews instead passed `<prompts>/tool.md`, a file, and one that exists in no prompt bundle. Every call therefore failed while canonicalizing the base, with "Failed to canonicalize base path: No such file or directory". The tool is how a stage reaches a guide the pre-screen did not select: any of the subsystem guides, severity.md, false-positive-guide.md. Losing it costs a review the guidance it went looking for, and it fails quietly, because a tool erroring is indistinguishable from a model deciding not to call it. Pass the directory. The test pins both directions, since the failure only shows up as an absence. Signed-off-by: Kees Cook <kees@kernel.org>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The read_prompt tool resolves the name it is given against a base path, so that base has to be the prompt directory: ToolBox is constructed with one everywhere else, and its test reads "technical-patterns.md" relative to it. Worker-run reviews instead passed
<prompts>/tool.md, a file, and one that exists in no prompt bundle. Every call therefore failed while canonicalizing the base, with "Failed to canonicalize base path: No such file or directory".The tool is how a stage reaches a guide the pre-screen did not select: any of the subsystem guides, severity.md, false-positive-guide.md. Losing it costs a review the guidance it went looking for, and it fails quietly, because a tool erroring is indistinguishable from a model deciding not to call it.
Pass the directory. The test pins both directions, since the failure only shows up as an absence.