Update ElevenLabs agent skills from a merged weekly changelog in elevenlabs-dx, then open a pull request in elevenlabs/skills. Trigger after a changelog merges to main on elevenlabs-dx, or when asked to update skills from changelog YYYY-MM-DD.
update-skills-from-changelog is an internal maintenance automation for ElevenLabs' agent-skills repository. It solves the upkeep problem of keeping evergreen skill documentation in sync with product changes: after a weekly changelog merges to main in the elevenlabs-dx repo, this skill discovers what changed, decides which skills are affected, and applies targeted, timeless documentation edits, then opens a pull request in elevenlabs/skills.
The workflow is disciplined. It resolves a changelog date, fetches the merged changelog via the gh CLI (base64-decoding the file contents), and applies a relevance filter mapping changes to specific skills such as text-to-speech, speech-to-text, agents, music, and setup-api-key. Crucially, before editing, it verifies every field, parameter, schema, enum, endpoint, model ID, or SDK method against canonical source documentation rather than inferring from changelog wording; anything without enough schema detail is deferred to a Needs Manual Authoring section. It then applies only edits that pass a fit gate, self-checks them, creates a skills-update branch, commits, pushes, and opens a PR. It explicitly does not read GitHub issues and does not touch skills the changelog does not implicate.
The target user is effectively the ElevenLabs documentation-maintenance workflow itself, run as automation or on demand. Its actions are standard developer operations — read a changelog, edit markdown, open a PR — with a strong emphasis on accuracy and minimal, well-scoped changes.
After a weekly changelog merges to main in the elevenlabs-dx repository, or when explicitly asked to update skills from a changelog for a given YYYY-MM-DD date.
It uses the gh CLI to pull the merged changelog markdown from elevenlabs-dx main and base64-decodes it to a temp file, optionally also reading the published changelog page.
It verifies every documented field, parameter, schema, enum, endpoint, model ID, and SDK method against canonical source docs, never infers schemas from changelog wording, and defers under-specified items to a Needs Manual Authoring section.
No. It applies a relevance filter and edits only skills implicated by the changelog, skipping internal/admin API changes, backward-compatible renames, and unrelated dashboard or pricing changes.
A skills-update/YYYY-MM-DD branch with committed, self-checked, evergreen edits and an opened pull request in elevenlabs/skills, or a clean stop with no PR if no skills are affected.
Quick Setup:
.claude/skills/Repository
elevenlabs/skills