feat(tri): tri skill claims -- the figures in the knowledge base, and what re-takes them - #3018
Merged
Conversation
… what re-takes them 409 numbered sections; 121 state a figure in their heading; 40 name a command a second reader could run; 116 are free to go stale. That rot surface has never been counted here. The population and the anchor rule are the ones tri issues numbers and tri issues dated already use, pointed at a second subject -- an address is not a count in a skill either, and a section saying "I published 268 and it was wrong" is history rather than a claim about the tree. But the SUBJECT had to change with it. Read from section BODIES the same rule reports 404 of 409, which is a matcher describing its input: a section is twenty-five lines of prose about numbers. Read from the heading, where the claim is actually made, it reports 121. A rule written for a one-line claim does not transfer to a page of argument by being pointed at it. Also names the three tracked SKILL.md files outside .claude/skills that no command here reads, rather than leaving them out in silence. Refs #2994 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
cargo fmt sorted forty mod lines this change never touched -- the trap already written down for a crate root, met inside the one file the change does edit. The diff here is now the two lines that wire the subcommand. Refs #2994 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…er inside the edited file 446: the first count of this file's own rot surface. 409 sections, 121 stating a figure in the heading, 40 naming a command, 116 free to go stale. Read from bodies the same rule reports 404 of 409 -- a matcher describing its input. A second probe reports 123 and the two-section gap is NOT resolved; both are stated. 447: cargo fmt sorted forty mod lines inside the file the change edits, so git status looked correct and only the diff shape showed it. Refs #2994 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Contributor
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-03 11:31:02 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
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.
The measurement
.claude/skills/ci-gates/SKILL.mdis 11,389 lines and 409 numbered sections, and until now nothing had ever asked how many of its own claims a second reader could re-take.A figure with no command beside it is not wrong — it is unre-takeable by anyone but its author. That is the rot surface, and it is the first time it has been counted here. The three tracked
SKILL.mdfiles outside.claude/skillsare named in the output rather than left silent: a count that quietly excludes part of its subject is the defect this page keeps recording.Reuse, and the part that did not transfer
The population rule and the anchor rule are the ones
tri issues numbersandtri issues datedalready use, pointed at a second subject — an address is not a count in a skill either (#2994,Wave Loop 369,w699are all over these headings), and a section that says "I published 268 and it was wrong in both directions" is history rather than a claim about the tree.But the subject had to change with the rule. Read from section bodies the same matcher reports 404 of 409 — a matcher describing its input, since a section here is twenty-five lines of prose about numbers. Read from the heading, where the claim is actually made, it reports 121.
Not resolved, and said rather than smoothed over
A throwaway probe over the same headings reports 123. Four attempts to locate the two-section gap failed — the comparison kept matching on truncated titles rather than on sections — and the shipped number is the one the tested matcher gives. Two readings of one population disagree by 2, the direction is not established, and §421 says what to do about it: report both and say which one the tool uses. That is in §446 and in this PR rather than absent.
Controls
cargo test -p triAlso in here
cargo fmtsorted fortymodlines inmain.rsthat the change never touched — the trap already written down for a crate root, met inside the one file the change does edit, sogit statusshowed exactly the two files I meant to touch and onlygit diff --statgave it away (45 changed, 20 deletions, for a two-line wiring change). Reverted; the wiring is now 9 insertions, 0 deletions. §447.Refs #2994
🤖 Generated with Claude Code