link-fixer — sweeps published stories for dead links and stale dates

Crawls every story's sources weekly, flags 404s / 5xx / domain expirations, and proposes archive.org wayback URLs as replacements via PR.

cp .claude/agents/link-fixer.md ~/.claude/agents/

What this agent does

Source links rot. arXiv URLs occasionally migrate, blog posts disappear, news sites paywall older content. link-fixer runs weekly, catches the breakage, and proposes Wayback Machine URLs as in-place replacements without changing the source title.

Repair priorities

1. Wayback snapshot at or near publish date — the version of the source the story actually cited
2. archive.today snapshot — secondary fallback, useful for paywalled news
3. Replacement on the same domain — if the publisher moved a URL but the article still exists
4. Mark as [archived: true] if no snapshot exists — preserves the citation but tells readers the link is dead

What it does NOT do

• Doesn’t change the source title field (that’s the citation, not the URL)
• Doesn’t delete sources — that would break historical attribution
• Doesn’t auto-merge — every change goes through the operator’s PR review

Edge case: domain transfer

When a publisher’s domain transfers (e.g., theinformation.com → pro.theinformation.com), the bot infers the rewrite rule from 3+ examples on the same domain in the same week and applies it consistently rather than treating each as an independent fix.