agent-ix/dev-team
Team planning, leadership, coding, review coordination, feature feedback, and program status workflows.
Run the coding lane for one team as its long-lived code lead. Pull ready work from the team's tracker plan in order, dispatch spec authors, coders and reviewers as subagents in fresh contexts, fix every finding inside the original PR, verify and merge reviewed work, keep tracker state true, and keep landing for hours or days without waiting on the owner. Use when invoked as /code-lead <team or project> or told "you are code lead for X" / "you are MAIN TEAM".
Run a ticket's coding stage as the dedicated coder role. Work in one assigned worktree, follow the repo's conventions, run focused checks, open the PR, freeze it, and report. Fix review findings inside the same PR. The aggregate gate runs once before merge. Use when invoked as /coder or dispatched by /code-lead.
Run an acceptance check on a delivered feature against its ticket's acceptance criteria and return a per-criterion PASS/FAIL verdict with evidence. Use when a feature is claimed done, before closing its ticket, or when asked to accept, verify or UAT a feature. For open-ended experience feedback use feature-feedback.
Give structured works / friction / gaps feedback on a feature you have just used, as a stakeholder or UAT tester. Use after trying a script, tool, skill or workflow, or when asked to test something and report back. To request something new use feature-request; for a pass/fail acceptance verdict use feature-check.
Write a request for a new capability as a Linear ticket that another agent can build and a stakeholder agent can later accept. Use when you hit something the tooling cannot do, want new functionality, or are asked to file a feature request. For feedback on an existing feature use feature-feedback; to verify a delivered feature use feature-check.
Run a repeatable, evidence-backed review across a scoped program, reconcile project state, assess value, and prepare a concise HTML report.
Answer one bounded, read-only question about one or more repositories using origin/main evidence. Use for measured repository inventories, shared-crate adoption, feature existence, or repository purpose; do not use for implementation or code review.
Run a PR's review stage as the dedicated reviewer role — run every applicable review method (code-review, rust-review, quoin gap-analysis, quoin spec-review and its sub-analyses) in one pass, let each write its own quire-validated SpecReview artifact, and post findings and later dispositions to the project's ticket tracker (Linear by default) with machine-readable markers so review data survives outside chat and each comment can be mined on its own, with no git join needed. Use when invoked as /reviewer, when dispatched as the reviewer subagent for a PR (for example from /code-lead), or when a review needs to be logged to the tracker instead of only the PR.
Own the STD spec-gap queue. Resolve gaps that block implementation first, value-test the rest, and cancel tracking ceremony. Use when dispatched by /code-lead; spec work does not need or hold a build lane.
Report evidence-backed status for a Linear initiative, project, or multi-repository program, including completed work, blockers, architecture and quality concerns, and the next owned action.
Watch the agents of one or more teams on an interval and unstick them by message. Reads each agent's real activity with ix-board agents, never asks for status, and messages the idle, stuck or blocked agent, its code-lead, its planner or the user. Use when invoked as /team-minder <team(s)> [interval], or when asked to keep an eye on a team's agents, find idle or stalled leads, or run a minder loop.
Turn an idea or change request into a structured Linear program, project, and ready tickets with dependencies, specification and architecture gates, then hand execution to a code lead or coder. Also review an existing team focus.
Independently evaluate whether a team's planner, code lead, coders, reviewers and minder are converging on their gate tickets, on a schedule, like a sprint retro. Takes metric snapshots with ix-board retro, compares them with the previous snapshot and a baseline, scores the role rubrics, names the repeating loops and messages the planner, code lead or user with owned actions. Use when invoked as /team-retro <team> [interval], or when asked to find out why a team is busy but not finishing, to run a retro, or to set up a recurring health check.