medium — read & publish on Medium via your own cookies

Drives the user's real Medium account through the same internal web API the site uses (Medium retired its public write API in 2023), authenticated by the login cookie they captured with the ACE extension. No browser, no third-party deps — just urllib.

The connector injects the cookie jar as an env var:

MEDIUM_COOKIES — a JSON array of cookies (sid, uid, xsrf). Secret — never echo or print it. The CLI echoes the xsrf cookie as the x-xsrf-token header on writes for you. If the captured jar has no xsrf cookie (common), the CLI mints one automatically before writing, so publishes no longer fail with "Missing xsrf token".

CLI

The skill ships scripts/medium.py — self-contained, stdlib only.

# $SKILL_DIR can point at another skill loaded this turn — anchor on our own
# script, and re-run this at the top of every Bash block (fresh shell each time).
MED="$SKILL_DIR/scripts/medium.py"; [ -f "$MED" ] || MED=$(find /tmp -maxdepth 8 -path '*/skills/*/scripts/medium.py' 2>/dev/null | head -1)
[ -f "$MED" ] || { echo "medium script not found (SKILL_DIR=$SKILL_DIR)" >&2; exit 1; }
python3 "$MED" whoami                     # who is logged in
python3 "$MED" articles --limit 20        # my posts + clap/response stats
python3 "$MED" article <post-id>          # one post's details

Verify the connection first

MED="$SKILL_DIR/scripts/medium.py"; [ -f "$MED" ] || MED=$(find /tmp -maxdepth 8 -path '*/skills/*/scripts/medium.py' 2>/dev/null | head -1)
python3 "$MED" whoami
# → {"user_id": "...", "name": "...", "username": "..."}

On an auth error the cookie is expired — have the user reconnect at https://studio.acedata.cloud/console/connectors. Do not loop-retry.

Publishing — GATED (dry-run unless trailing `--confirm`)

publish writes to the user's real account. Content is Markdown (converted to Medium paragraph blocks: #/##→heading, ###→sub-heading, >→quote, ```→code, -/*→bullet list, 1.→numbered list, tables→aligned code block). Inline bold, italic, code, and [links](url) render as real Medium formatting (clickable links included). Bare https://… URLs in text are auto-linked too (URLs inside code spans are left untouched). Without a trailing --confirm it dry-runs. --confirm is honored only as the last argument. Always show the dry-run, get an explicit "yes", then re-run.

MED="$SKILL_DIR/scripts/medium.py"; [ -f "$MED" ] || MED=$(find /tmp -maxdepth 8 -path '*/skills/*/scripts/medium.py' 2>/dev/null | head -1)
python3 "$MED" publish --title "Title" --content-file a.md                       # dry-run
python3 "$MED" publish --title "Title" --content-file a.md --draft-only --confirm   # private draft
python3 "$MED" publish --title "Title" --content-file a.md --confirm                # PUBLIC story

Publishing is Medium's multi-step editor flow (new-story → write deltas → publish). --draft-only stops at the draft (visible only at the user's /me/stories/drafts). Default to --draft-only unless the user asked to go live.

Images

publish automatically uploads each external markdown image (![](url)) to Medium and inserts it as a real image block — Medium has no markdown-image syntax, so this is the only way images render. Pass --no-rehost-images to degrade images to link-only paragraphs. An image that fails to upload falls back to a link paragraph (never blocks the post).

Gotchas

This is the user's real Medium account. Confirm before any publish.
Markdown→Medium conversion covers headings, quotes, fenced code, bullet/ numbered lists, images, and inline markups (bold/italic/code/links). Medium's editor has no table element: a narrow markdown table renders as an aligned monospace code block, while a wide one (long cells / URLs, which would wrap into an unreadable grid) is rendered as per-row records — the first column bolded as a lead-in, the remaining columns as header: value bullets.
Medium sits behind Cloudflare; an occasional 403/429 is transient — the CLI auto-retries once after a short pause. A persistent 403 means the cookie is genuinely expired (reconnect).
Never print MEDIUM_COOKIES — it is full account access.
ToS: acts only on the user's own account with their own captured cookie.

Record the output

After you successfully publish and obtain the live result URL, call the built-in publish_artifact tool ONCE so the user can track this deliverable in My Outputs:

publish_artifact(kind="article", channel="medium", title="<title>", url="<the REAL returned URL>", status="delivered")

Use the real returned URL — never fabricate one. Call it once per published item, only after delivery is confirmed; skip it (or use status="failed") if publishing failed. See _shared/artifacts.md.

medium