Add Customer To User List

Single source of truth: maintain this skill under .agents/skills/add-customer-to-user-list/ only. Claude and Cursor load projected copies under .claude/skills/add-customer-to-user-list and .cursor/skills/add-customer-to-user-list.

Add a customer to the users list

Adds or updates one row in the public adopters table on /users.

Do not write a customer story in this skill. If the user wants a full /users/<slug> case study, use customer-story-setup first, then come back here to link it.

Canonical file

The handbook imports the same table (<AdoptersTable />). Editing the MDX file updates both pages. Do not duplicate the table elsewhere.

Before writing any file

Collect what you can from the request. Ask only for what is still missing:

1. Company name — public marketing name
2. Company website — used in the Company column
3. Use case — optional; write one if a user story or external link exists
4. External reference URL — optional public source (blog, job post, DPA, talk)

Then always check for a user story. Do not skip this step.

Always check: is there a user story?

Search the working tree and origin/main (this branch may be behind):

• content/customers/meta.json → pages (skip "index")
• content/customers/*.mdx — match company name, legal name, or obvious slug variants (hugging-face vs huggingface, merckgroup vs merck)
• Frontmatter showInCustomerIndex and quoteCompany if the filename is unclear

The public URL is /users/<slug> where <slug> is the MDX filename without .mdx.

Never write Langfuse Customer when a user story exists. Prefer the user story over an external link when both exist.

External reference links

When the user (or a public source they pointed at) provides a URL and there is no user story:

1. Open the page and read enough to understand the Langfuse mention.
2. Middle column: one short use-case phrase (sentence case), focused on what they do with Langfuse or their AI product — not a full company bio.
3. Last column: a short link label plus the URL, matching existing rows.

Link-label examples already in the table:

Blogpost, Tech Blogpost, AWS Blogpost, LinkedIn Post, Job Description, Case Study, Client Story, Documentation, Release Notes, Research Paper, Source Code, Conference Talk, Conference Deck, Privacy Policy, Data Processing Addendum, Subprocessor List, IT White Paper, Technical Memorandum, Open Source Catalog, OSS LLMOps Stack, Partner Newsletter

Pick the closest label. Include a date only when it helps (Blogpost from 12/23/2025).

Always include a Wayback Machine archive when the Reference cell is an external URL. User-story links (/users/<slug>) and Langfuse Customer do not need one.

1. Look up an existing 200 snapshot. Always percent-encode the URL with curl -G --data-urlencode (or equivalent). A raw ?url= query treats & in the page URL as another CDX parameter and can miss snapshots.

   curl -sS -G "https://web.archive.org/cdx/search/cdx" \
     --data-urlencode "url=<URL>" \
     --data-urlencode "output=json" \
     --data-urlencode "filter=statuscode:200" \
     --data-urlencode "limit=1"

2. If CDX is empty, request a snapshot and wait for it. Do not skip this step or ship the row without an archive. Percent-encode reserved characters in the page URL (?, &, #) so Save Page Now receives the full URL.

   curl -sS -I "https://web.archive.org/save/$(python3 -c 'import urllib.parse,sys; print(urllib.parse.quote(sys.argv[1], safe=":/"))' '<URL>')"

   Re-query CDX until it returns a 200 row. The Save Page Now response may be 5xx even when a capture is stored — trust CDX plus a fetch of the archive URL.

3. Confirm the snapshot contains the Langfuse mention (or the cited source content), then use:

   https://web.archive.org/web/<timestamp>/<URL>

[Tech Blogpost](https://example.com/post) ([Archive](https://web.archive.org/web/20260304001459/https://example.com/post))

If Wayback cannot capture the page after a retry (common for some LinkedIn or login-walled URLs), use an archive.ph snapshot the same way neighboring rows do. Say so in the PR. Do not omit the archive.