Skip to main content

archive-file

What it does​

archive-file moves one stale file into your repo's archive folder, stamps its frontmatter with just enough metadata to explain itself later (created, summary, archived/archived_date, and optional parent/children/related links), records the move in a running archive log, and flags — without touching — any other page that still links to it.

It is a minimal v1 primitive. It does not decide a file is stale for you, and it does not go rewrite every other page that referenced it. Both of those are explicitly out of scope for now.

Locating the archive folder and log​

Neither location is hardcoded. The skill scans the repo for a folder literally named _archive, and for a file named archive-log.md alongside wherever the rest of the wiki's index/log files live — the same scan-for-a-marker approach resolve-ai-paths uses for session/prompt folders. First run in a repo with neither yet: it proposes sensible defaults and asks before creating anything.

What goes on the file​

  • created — pulled from the file's first commit in git history where possible, otherwise asked
  • summary — drafted from the file's own content and shown to you to confirm or edit
  • archived / archived_date — the lifecycle stamp
  • parent / children / related — optional, only added if you say they apply

Every existing frontmatter field on the file is left untouched; this only adds fields, never rewrites others.

Before it moves anything​

It asks whether the file should go through something else first — captured as a task, ingested, cleaned up, whatever fits. If you're not sure what's available, it points you at /ask-alex rather than keeping its own list of candidate skills, which would just go stale the next time a skill is renamed or added.

What it won't do​

  • Auto-rewrite other pages' links to the file it just archived — it lists them, you fix them
  • Detect staleness on its own — you decide a file is stale; this skill only acts on that decision
  • Invent a frontmatter convention if your repo doesn't already document one — it asks rather than guessing

It's working if​

  • The archived file is self-describing: anyone opening it cold can tell what it was, when it was written, and when/why it was archived, without cross-referencing anything else.
  • The archive log has one new entry pointing back at the file.
  • Nothing else in the repo silently broke — any dangling reference was surfaced, not swept under the rug.