Tighten prose-writing prompt step to re-check CLAUDE.md per sentence#39
Merged
Merged
Conversation
…tence PR #37's prose changes drifted into changelog-style framing in eight places ("now use the same path", "currently only supports", "legacy daemon-wallet path", "until upstream X exposes Y", "preserves pre-0.6.1 behavior", "not yet supported"). Every one of these violated rules that already exist in CLAUDE.md — the Style guide explicitly bans time-relative wording, the Voice-and-tone section mandates present tense, and the "do not narrate the evidence" rule asks for direct statements of fact. The general CLAUDE.md pointer in step 4.7 was easy to skim past during the audit/write loop. Tighten step 4.7 to call out the specific failure mode the routine keeps hitting: the audit gives the model a diff between two SHAs, but the page describes the state of the surface at head_sha. Translating "what changed" into "what is" is exactly where the model slips. The new wording names the relevant CLAUDE.md sections explicitly and gives one concrete bad-vs-good example without duplicating the rule itself. No rule duplication; the Style guide remains the source of truth. https://claude.ai/code/session_019cz1M7pXuXFQykMZxfdXHm
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
PR #37's prose changes drifted into changelog-style framing in eight places — "now use the same path", "currently only supports", "legacy daemon-wallet path", "until upstream X exposes Y", "preserves pre-0.6.1 behavior", "not yet supported". Every one of these violated rules that already exist in CLAUDE.md — the Style guide bans time-relative wording, the Voice and tone section mandates present tense, and the "do not narrate the evidence" rule asks for direct statements of fact. The general CLAUDE.md pointer in step 4.7 was easy to skim past during the audit/write loop.
This PR tightens step 4.7 to call out the specific failure mode the routine keeps hitting: the audit gives the model a diff between two SHAs, but the page describes the state of the surface at
head_sha. Translating "what changed" into "what is" is exactly where the model slips. The new wording names the relevant CLAUDE.md sections explicitly and gives one concrete bad-vs-good example.No rule duplication — the Style guide remains the single source of truth. The prompt just re-points at it more emphatically with the explicit failure-mode framing.
Test plan
https://claude.ai/code/session_019cz1M7pXuXFQykMZxfdXHm
Generated by Claude Code