Repository navigation
Extend the source catalog across more departments - #42
Merged
Merged
Conversation
Research across the twelve uncovered departments, then verification of every result. 50 sources to 181; 36 skills covered to 144. The discipline mattered more than the breadth. The instruction throughout was that a source only counts if it settles a disagreement between two competent practitioners — a regulator, primary law, a standards body, or a specification — and that a short list beats a long one. So a large number of skills deliberately carry nothing. Nobody adjudicates how to write a good line, run a discovery interview, choose a visual register, debug, or decide when work is done, and a confident blog post attached to one of those would lend false authority rather than add any. Where a skill is part craft and part constraint the split is stated: law bounds what a claim may say and how a cancellation flow must behave, and settles nothing about persuasion or taste. The licensing picture is the reason the catalog records a license class at all. Of 181 sources, 122 are quotable and 59 are read-and-cite — copyrighted, free to read, not free to reproduce. The largest quotable block is US federal work, which is public domain by statute. The standards most often assumed open are not: the quality, continuity, AI-management and SQL standards are sold, the payment security standard and the consensus hardening baselines need an account, and the trade terms that decide where risk transfers between buyer and seller are sold too. Each entry says which, in the imperative, beside the link. Verification found no dead links across 213 URLs. Thirty are refused by this network — government and code-hosting domains that bot-filter — and the weekly workflow runs from elsewhere, which is what it is for. One host was unreachable entirely at the end of the run having answered earlier the same day. Two defects surfaced while checking, both in the checkers rather than the content. The spelling check scanned URLs, and --fix rewrote them in place. A British spelling inside a link therefore became a broken link, and the build went green afterwards. A UK regulator whose own path contains one exposed it. URLs are now masked before matching, so prose in a sentence containing a link is still checked while the link is left alone; the repair was verified in both directions. The provenance waiver for license family names excluded the catalog itself, on the reasoning that entries carry a slug rather than a family's prose name. That stopped being true when a license became a source: a font license is the authority on whether a typeface may be embedded, so its title is the name. The waiver now covers the catalog and what is generated from it, and remains name-only — confirmed by pasting real license text in and watching it still fail. All fourteen checks pass. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FaQP2WCqeFYH7s9pi1CJ5u
cbrock84
changed the title
cbrock84
marked this pull request as ready for review
September 16, 2026 21:53
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
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.
Research across the twelve uncovered departments, then independent verification of every result.
50 sources → 181. 36 skills covered → 144.
The discipline mattered more than the breadth
The rule throughout: a source counts only if it settles a disagreement between two competent practitioners — a regulator, primary law, a standards body, a specification. Not a blog, not vendor content marketing, not thought-leadership.
So a large number of skills deliberately carry nothing. Nobody adjudicates how to write a good line, run a discovery interview, choose a visual register, debug, or decide when work is done. Attaching a confident-sounding source to one of those would lend false authority rather than add any.
Where a skill is part craft and part constraint, the split is stated rather than blurred. Law bounds what a claim may assert and how a cancellation flow must behave; it settles nothing about persuasion or taste.
Why the license field exists
Of 181 sources, 122 are quotable and 59 are read-and-cite — copyrighted, free to read, not free to reproduce.
The largest quotable block is US federal work, public domain by statute. The standards most often assumed open are not:
Each entry says which, in the imperative, beside the link.
Verification
No dead links across 213 URLs. Thirty are refused by this network — government and code-hosting domains that bot-filter — which is what the weekly workflow, running from elsewhere, is for. One host was unreachable entirely at the end of the run having answered earlier the same day.
Worth noting: the research reported a number of URLs as unverifiable or 403. Checked from here, every one of them resolved 200. Those blocks were the researcher's network, not the sites. I verified independently rather than inheriting either the pass or the fail.
Two defects found — both in the checkers, not the content
The spelling check scanned URLs, and
--fixrewrote them in place. A British spelling inside a link became a broken link, and the build went green afterwards. A UK regulator whose own path is/for-organisations/exposed it. I reproduced the corruption before fixing it:URLs are now masked before matching, so prose in a sentence containing a link is still checked while the link itself is left alone. Verified in both directions.
The provenance waiver for license-family names excluded the catalog itself, on my earlier reasoning that entries carry a slug rather than a family's prose name. That stopped being true the moment a license became a source — a font license is the authority on whether a typeface may be embedded, so its title is the family name. The waiver now covers the catalog and what is generated from it, and remains name-only: confirmed by pasting real license text into a waived file and watching it still fail.
Verification
./scripts/check-all.sh— all fourteen green.🤖 Generated with Claude Code
https://claude.ai/code/session_01FaQP2WCqeFYH7s9pi1CJ5u
Generated by Claude Code