v0.2.0

Added

• The terminal card names itself. One line at the foot of the card reads
  agent-wrapped.dev. A screenshot loses every link around it, and that line is
  what tells the next reader what they are looking at. It is a printed string,
  not a network call — the no-telemetry rule is unchanged.
• Machine-readable documentation. agent-wrapped.dev now serves /index.md
  (the whole page as Markdown), /llms.txt, /robots.txt and /sitemap.xml, and
  the page carries SoftwareApplication and FAQPage structured data. An agent
  asked about Claude Code usage statistics can fetch plain text instead of
  scraping 13 KB of inline CSS.
• A questions section on agent-wrapped.dev, answering where Claude Code keeps
  its transcripts, how long it keeps them, how to change cleanupPeriodDays, and
  how the card differs from /insights.
• npm keywords went from seven to eighteen, adding the terms people actually
  search — anthropic, agent, token-usage, analytics, year-in-review.

v0.1.2

Added

• Shareable image. The --html card now has a Save as image button that
  renders a 1200×630 PNG in your browser and saves it locally. Nothing is
  uploaded — the image is drawn on a canvas and saved through an object URL, and
  only aggregate numbers ever reach the page.

Changed

• The HTML card is written to ~/.agent-wrapped/card.html instead of a
  system temp path. Everything the tool writes now lives in one directory you can
  find again. If you scripted the old temp path, update it; the stale file is
  harmless and your OS will clear it.

Fixed

• The "open it with" hint was macOS-only. Linux and Windows users were told to
  run open, which does not exist there. It now prints xdg-open or start as
  appropriate.

v0.1.1

Fixed

• Package metadata: homepage now points at https://agent-wrapped.dev.

Notes

• First release published through GitHub Actions with npm trusted publishing, so
  this is the first build carrying a provenance attestation. No functional change
  to the tool itself.

v0.1.0

Initial release.

Added

• A wrapped card for how you work with Claude Code, read from your own
  transcripts: read:write ratio, cache share, subagent split, repo concentration,
  skill mix, tool mix, turn depth, streaks and rhythm, and usage-limit hits.
• --since to limit the window, which is also faster than a full run because
  transcripts older than the cutoff are never opened.
• --html for a self-contained page, --json for raw stats.
• Snapshots: each full run saves a ~1 KB summary to ~/.agent-wrapped/snapshots/
  and compares against the previous one, so you can keep a year of history
  without keeping a year of transcripts.
• A size-aware retention warning that measures your own growth rate rather than
  telling everyone to keep a year of transcripts.