Offered by Dead Brands
Every link on a stream page checked before broadcast — dead ones named, fixed, or flagged.
Use it over MCP: connect to https://mcp.deadbrands.co/hub/mcp and call skills_get with slug dead-link-archaeology.
# Dead-Link Archaeology for Stream Pages ## Purpose Concrete use case — Used by: the stream-page maintainer bot. Task: sweep every link on a game TV/stream page before it goes live, so viewers never hit a dead page mid-broadcast. Each link gets a target, a status, and a last-checked time. It reports only what it actually fetched, and it never rewrites a page on its own — it hands the maintainer a fix list. ## Superpower Every link on a stream page checked before broadcast — dead ones named, fixed, or flagged. ## When to use - Before any stream page ships or re-ships, and after any page move or rename. - When a viewer or teammate reports a link that goes nowhere. ## When NOT to use - Do not use it to crawl sites outside the team's own pages without a reason, or to guess a replacement URL. - Do not auto-edit pages; this skill finds and reports, it does not publish fixes. ## Tools & APIs The stream page files themselves (existing, real); a link extractor and a plain HTTP status check (existing, real). Planned, not yet available: a link-report table (URL, status, first-seen-dead date) — until it exists, record those fields by hand. ## Inputs The page file or URL list to sweep; the expected target for each link (same site, stream page, or documented outside page); a freshness window for re-checks. ## Outputs A table of every link with its live status; a fix list of dead or redirected links with page and line; a clean bill for zero-dead-link pages. ## Example invocation > "Sweep the Asteroid Dodge stream page: list every link, fetch each target, and return the dead ones with page and line." ## Steps 1. Collect every href on the page, including links inside buttons and footers, and dedupe them. 2. Fetch each target once and record the status code and final URL after redirects. 3. Mark a link dead only on a failed fetch or a 404/410; mark redirects separately, never as dead. 4. Re-check each dead link once after a short wait, to rule out a momentary outage. 5. Hand the maintainer the fix list with page, line, and old target — plus a replacement only when a real one is known. ## Safety limits Never publish a page edit from this skill alone; a human or the page owner approves fixes. Never invent a replacement URL — an unknown replacement is reported as unknown. Never hammer a server; one check plus one re-check is the limit. ## Failure modes - **Redirect mistaken for dead:** a link that moves is reported as redirected with its new target, not as broken. Mitigation: record the final URL, not just the first status. - **Momentary outage mistaken for death:** a page down for a minute looks dead. Mitigation: the mandatory re-check before any fix list is sent. - **Hidden links missed:** links inside scripts or footers escape a naive scan. Mitigation: scan the raw page source, not just the rendered text. ## Test checklist - [ ] Every link on the page appears in the table, including footer links. - [ ] Dead, redirected, and live links are three separate categories, never merged. - [ ] Each dead link was re-checked once before being reported. - [ ] No replacement URL is suggested unless a real one was verified. - [ ] No page was edited by this skill; the fix list was handed on. --- *Trademarks and copyrights mentioned belong to their respective owners. Educational reference only; not affiliated with or endorsed by them.*