youtrack skill
Every operation this skill described is now covered by the youtrack skill, which drives
.agents/skills/youtrack/scripts/youtrack.ts through the repository-pinned Bun wrapper. Read that
skill's SKILL.md and use its commands.
Why it was replaced:
- The CLI resolves the token from
YOUTRACK_TOKENor the native OS credential store, so the token never has to be exported into the shell environment. - Bodies are piped over stdin instead of staged in
mktempfiles, so there is no extra write approval and no temp file to clean up. - Two reusable approval prefixes (
… youtrack.ts readand… youtrack.ts write) cover the whole surface, and the read/write split makes mutations an explicit decision. - Ids, link directions, tag names, and custom field
$types are resolved by the CLI rather than hand-assembled per call.
Command mapping
| This skill's recipe | Replacement |
|---|---|
| Get an issue | read issue get <ID> |
| Search issues | read issue search --query <query> [--top n] |
| List saved queries | read saved-query list |
| Create an issue | write issue create --project <SHORT_NAME> --summary <text> [--description-file -] |
| Update summary or description | write issue update <ID> [--summary <text>] [--description-file -] |
| Apply a command (state, priority, assignee) | write issue state <ID> --value <state>, write issue set-field <ID> --name <field> --value <value> |
| List comments | read issue comments <ID> |
| Add a comment | write issue comment <ID> --text-file - |
| Update or delete a comment | write comment update <ID> --comment <CID> --text-file -, write comment delete <ID> --comment <CID> |
| List or search tags | read issue tags <ID>, read tag list --query <text> |
| Add or remove a tag | write issue tag <ID> --name <tag>, write issue untag <ID> --name <tag> |
| Read links | read issue links <ID> |
| Create or remove a link | write issue link <ID> --type <phrase> --target <ID>, write issue unlink … |
| List link types | read link-type list |
| List or add work items | read issue work-items <ID>, write issue log-work <ID> --duration '2h 30m' |
| Look up users | read user me, read user search --query <query> |
| Inspect project custom fields | read project fields <SHORT_NAME> |
| List projects | read project list |
| Read attachments | read issue attachments <ID>, read issue attachment <ID> --name <name> [--out <path>] |
Not covered by the CLI: uploading attachments, group lookup, agile boards, and project administration. Those need a separately approved API call.
The old advice to pass project.id as an internal numeric id (22-1758 for JEWEL) is obsolete —
the CLI creates issues with the project short name.
Jewel specifics worth keeping
- The JEWEL project requires the Type and State custom fields on create. Priority is also required, but only Jewel team members may set it.
- For Jewel workflows around ticket state transitions and releases, see
pr-guide.md(PR conventions including YouTrack issue references in commit messages) andreleasing-guide.md(release process that involves YouTrack issue tracking).