Issue Trackers (Jira & Linear)
Accelerator integrates with two remote issue trackers — Jira and
Linear — through one shared pattern. Read skills (search, show)
auto-trigger on natural-language phrasing; write skills are slash-only
and display a payload preview that you must explicitly confirm before
any change reaches the tracker. Each integration keeps a team-shared
catalogue (committed) alongside gitignored per-developer credentials,
and treats the presence of an external_id on a work item as the signal
that it is synced to the remote. Building on that signal,
sync-work-items
reconciles meta/work/ against the tracker in batch. Select the active
tracker with the work.integration key (jira or linear); when
unset, the work-management skills stay local with no external API calls.
Both integrations ship the same eight skill shapes:
| Shape | Jira | Linear |
|---|---|---|
| Initialise | init-jira | init-linear |
| Search | search-jira-issues | search-linear-issues |
| Show | show-jira-issue | show-linear-issue |
| Create | create-jira-issue | create-linear-issue |
| Update | update-jira-issue | update-linear-issue |
| Comment | comment-jira-issue | comment-linear-issue |
| Transition | transition-jira-issue | transition-linear-issue |
| Attach | attach-jira-issue | attach-linear-issue |
Run the init skill once per project before the others: it verifies
credentials and persists the team-shared catalogue under
.accelerator/state/integrations/.
The Jira skills target a Jira Cloud tenant. Add the shared site setting
to .accelerator/config.md and personal credentials to the gitignored
.accelerator/config.local.md:
# .accelerator/config.md — commit this---jira: site: your-subdomain # e.g. "atomic-innovation" for atomic-innovation.atlassian.net---# .accelerator/config.local.md — do not commit---jira: email: you@example.com token_cmd: "op read op://Work/Atlassian/credential" # any password-manager command---The default project key reuses work.default_project_code; set
work.integration: jira to enable auto-scoping. See /configure help
for the full credential resolution chain and token_cmd examples.
Jira Cloud stores rich text as Atlassian Document Format (ADF);
Accelerator converts bidirectionally — reading skills render ADF to
Markdown, writing skills convert your Markdown to ADF. The
init-jira state
cache (field catalogue, project list, site metadata) is
version-controlled and team-shared, so teammates don’t need to re-run
init.
The Linear skills talk to the Linear GraphQL API directly
(Markdown-native — no ADF conversion). Auth is token-only — no site
or email. Put a Linear personal API key in the gitignored
.accelerator/config.local.md:
# .accelerator/config.local.md — do not commit---linear: token_cmd: "op read op://Work/Linear/credential" # or linear.token: lin_api_…---.accelerator/config.local.md must be mode 0600 or stricter.
init-linear
fixes the workspace to a single team and caches that team with its
workflow states — catalogue.json is committed and team-shared,
viewer.json is gitignored and per-developer. Set
work.integration: linear to enable auto-scoping.
Per-skill flags, sub-actions, and behaviours are documented on each skill’s reference page linked in the table above.