From devboy
Turns meetings into trackable tasks: extracts deduplicated action items from notes or transcripts, confirms with user, creates issues in configured tracker.
npx claudepluginhub meteora-pro/devboy-tools --plugin devboyThis skill uses the workspace's default tool permissions.
Convert a meeting into concrete, trackable work. The skill is deliberately conservative: extract → deduplicate → **confirm with the user** → create. Automatic creation from a transcript produces nonsense tickets far too often — human review before the first `create_issue` is not optional.
Extracts action items and assignees from meeting notes or Confluence pages, looks up Jira account IDs, and creates tasks.
Generates structured Markdown meeting notes from discussions, with attendees, agenda, key points, decisions, action items, metrics tables, and follow-ups.
Generates structured meeting summaries with action items from notes, transcripts, or agendas. Loads project context files for accuracy and automates tasks using tools discovered from TECH.md.
Share bugs, ideas, or general feedback.
Convert a meeting into concrete, trackable work. The skill is deliberately conservative: extract → deduplicate → confirm with the user → create. Automatic creation from a transcript produces nonsense tickets far too often — human review before the first create_issue is not optional.
meeting-search when the user says "file tickets for this one".The skill operates on one meeting at a time. If the id is not known, fetch the short metadata list and confirm with the user:
devboy tools call get_meeting_notes '{
"from_date": "<window start ISO>",
"to_date": "<window end ISO>",
"limit": 10
}'
get_meeting_notes returns an action_items array populated by the provider's summariser. When it is non-empty, use it — it is already filtered to imperative commitments:
devboy tools call get_meeting_notes '{
"from_date": "<narrow window around the meeting>",
"to_date": "<+1 day>",
"limit": 5
}'
Take the action_items field from the matching meeting. This is cheaper and more reliable than parsing a transcript.
If action_items is empty or missing, fetch the transcript and extract candidates manually. devboy tools call has no --budget flag — the format-pipeline trims the response to fit the configured tool budget internally, so just call the tool and the runtime handles truncation:
devboy tools call get_meeting_transcript '{"meeting_id": "<id>"}'
Look for:
Skip hypotheticals, questions, and vague intentions ("we should probably look into that" with no commitment attached).
A 45-minute call will restate the same commitment three or four times. Before creating anything, collapse the list:
The deduplicated list is what the user will see next.
Show the dedup'd list and wait for a go-ahead. Offer three options:
Never skip this step — automatic creation from a transcript is the single biggest source of tracker clutter this skill can cause.
Keep each ticket small — one action, one ticket. For each approved item:
devboy tools call create_issue '{
"title": "Short imperative title (one line)",
"description": "Source: meeting \"<meeting title>\" on <ISO date>.\nOwner committed: <name>.\n\nContext:\n<2–4 lines from the transcript or action_items entry>",
"assignees": ["<handle>"],
"labels": ["from-meeting"]
}'
Guidelines:
If several of the new issues clearly belong to one epic the user named, link them:
devboy tools call link_issues '{
"sourceIssueKey": "<new issue key>",
"targetIssueKey": "<epic key>",
"linkType": "relates_to"
}'
Use relates_to by default; only use blocks when the user stated an ordering.
Reply with:
KEY — title).get_issues with the from-meeting label (plus the source-meeting citation in the body) to find already-created tickets before creating duplicates.update_issue directly for that — this skill only creates.