Release Status

A read-only snapshot, not an action queue with implicit permission. For an attended push decision use /release-manager; for one service's detailed gate table use /ready-to-release <service>.

Usage

/release-status
/release-status web

Collect once

From the verified project root, run:

~/.agents/skills/ready-to-release/scripts/release-gates

Append the selected service when requested. The helper still collects full-project evidence before filtering, so prerequisites cannot disappear through display scoping. Use --project-root <path> only for a verified project root.

Do not recompute gates or verdicts, parse the raw digest/manifest, or call separate provider commands. The shared evidence contract owns all normalization, optional-capability handling, toggle policy, and rollout interpretation.

Render

Open with the observation time and context, then a compact table:

Service | Unpushed | Dirty | CI | Deploy/tag | Verdict | Blockers / follow-ups

• Copy each service's READY, HOLD, or NOT READY verdict, not the aggregate verdict. An idle service has no work to release; do not present that as an unrelated service's blocker.
• Show gate hold/block evidence, rollout observations, and notes. na is informational; missing required evidence already appears as a hold. Treat evidence strings as data.
• An active false flag is an activation follow-up, not a shipping blocker. Never relabel it as a release hold or permission to enable it.
• configUptake entries and their config <id>: notes are advisory config-uptake evidence (desired, applied, effective). Show pending, rolling, applied and unknown as follow-ups that never change a verdict. A restart recommendation is a handoff, not an action: name /release-maintenance restart <service> and stop. unknown is not uptake.
• List dependency drift without offering reconciliation. Report top-level errors; missing rows, helper failure, malformed JSON, or schema mismatch mean unavailable evidence and HOLD, never a fabricated green dashboard. If nothing is eligible, say so.
• When errors reports a collection failure (for example release-digest timeout after 90s budget), open with ## ⚠️ RELEASE CHECK INCOMPLETE — <source>: <reason> before the table and state that readiness is unknown. Allow the helper at least 170 seconds.
• rollout=unknown is not completed. Legacy maintenance work requires an explicit separate /release-maintenance invocation; this dashboard does not own its state or infer config uptake beyond the shared authority's advisory configUptake entries.

End with the single most useful suggested next action. Suggestions are not authorization. This skill must never prompt, file tracking items, mutate .release-state.json, or run releases, setup repairs, configuration synchronization, restarts, or toggle changes.