docs(command): bound what a published verdict or observation may say (#141) - #142
Open
kevin-agrology wants to merge 1 commit into
Open
kevin-agrology wants to merge 1 commit into
kevin-agrology wants to merge 1 commit into
Conversation
…141) A record's text is copied verbatim to the PR author -- pr-watch relabels the grammar around it but never paraphrases a verdict -- and nothing in the command said so, or bounded it. A fourteen-round review posted multi-paragraph agrees and observations whose operative content was one clause, and the standing bounds in them repeated near-verbatim every round. Five sentences, all about the defect. Everything else goes in plain prose under the record: the gate reads it, compose-review does not ingest it (verified -- a prose line under an [agree:] composes clean and appears nowhere in the body). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MBFDiJrnWDv8JLda56bHiB
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #141.
The gap
Nothing in the command told the primary that a record's text is published verbatim to the PR author, or bounded how much of it to write. Downstream (
pr-watch) copies reviewer and adjudicator words exactly on purpose — it relabels the protocol's grammar but never paraphrases a verdict — so it cannot shorten what arrives. The result on a fourteen-round review was[agree:]records running to several hundred words, and[observation]blocks that were mostly standing session bounds,teeseeding mechanics, quarantine bookkeeping and commentary on the protocol itself, repeating near-verbatim every round.The rule
At most five sentences on a published record, every one about the defect — the verdict, the severity if it is being moved, the evidence that decides it, what the author must do.
Everything else goes in plain prose under the record: the human gate reads it, the composer never ingests it. Verified rather than assumed — a prose line under an
[agree:]composesrc=0and appears nowhere in the composed body, so the instruction does not walk the primary into a contract error.The diff also names what does NOT belong in a published record, however true (session bounds, dispatch and quarantine bookkeeping, round-bound and convergence commentary), and gives the observation rule its own test: it earns a place only if it tells the AUTHOR something about their change. A round with nothing for the author writes no observation.
The worked example in the diff is the real round-14
agreefrom agrology-field-digest#15, cut to five sentences with the bounds moved below the record.Checks
mutation-check: all 407 mutations caughtdocs-check: passesCommand prose only — no script, no schema, no version bump.
🤖 Generated with Claude Code
https://claude.ai/code/session_01MBFDiJrnWDv8JLda56bHiB