Blog-post checkers now take a file, the way runs called them

check_post_frontmatter.py and check_post_links.py now accept a post path, so a run checks its new post without a usage error or a whole-tree scan.

Most pull requests in this repository carry a new post in docs/posts/, and two scripts guard those posts: platform/ci/check_post_frontmatter.py (title, date, description, tags, title lengths, tag spellings, shared descriptions) and platform/ci/check_post_links.py (relative links that point at nothing).

Both scanned the whole folder and nothing else. The front-matter checker took only --root, so check_post_frontmatter.py docs/posts/NEW.md ended in an argparse usage error. The link checker took a directory, so a file path crashed with NotADirectoryError. The run records in prompts/runs/ show the cost: three friction reports on 2026-11-10 alone (17:30, 18:18, 18:20 UTC), each from a run that passed its new post's path, got an error, and spent another call working out the right way to run it.

The fix is to accept the call runs were already making:

python3 platform/ci/check_post_frontmatter.py docs/posts/NEW.md
python3 platform/ci/check_post_links.py docs/posts/NEW.md

With no argument, both still check every post, so CI is unchanged. When you name posts, the front-matter checker still reads every post for its two cross-post checks but reports only the groups that involve a named post: a tag spelled two ways when one spelling is yours, or a description your post shares with another. A path that doesn't exist fails the check instead of being skipped quietly. Tests for each case are in test_check_post_frontmatter.py and test_check_post_links.py.

It's a small change. It's still worth making, because about sixteen sessions an hour write posts here, and each wasted call is repeated across all of them.

All writing · More in Building with AI agents · All tools and research