1---2name: delivery-release-gate3description: Use `analysis-agent` for release decisions, `task-agent` for delivery artifacts, or `review-agent` for readiness on deployment, migration, rollback, or production risk. Skip local work with no release decision.4---56# delivery-release-gate78## Role910- **Analysis mode (`analysis-agent`):** Decide rollout, containment, and recovery behavior.11- **Task mode (`task-agent`):** Produce the accepted release artifact and rollback metadata.12- **Review mode (`review-agent`):** Judge artifact readiness against rollout and recovery criteria.1314## When To Use1516- production release; irreversible deployment or migration1718## Do Not Use1920- local implementation only21- no release decision2223## Required Inputs2425- release boundaries; rollback requirements and recovery constraints26- **Analysis mode (`analysis-agent`):** current release topology, blast radius, compatibility, and recovery evidence.27- **Task mode (`task-agent`):** accepted release-artifact decision with provenance, rollout, and rollback checks.28- **Review mode (`review-agent`):** release artifact with mixed-version and recovery evidence.2930## Professional Decision Rules3132- Prove release dimensions material to selected artifact, configuration, compatibility, migration, rollout, observability, and recovery risk.33- Select rollout, watch, approval, and containment from blast radius, reversibility, current controls, and policy.34- Test old/new coexistence or recovery against the actual artifact and material environment dimensions when triggered.35- Reserve destructive, production, privileged, or irreversible actions for explicit user authority.3637## High-Value Gotchas3839- A rollback command does not prove recovery.40- Clean deployment can miss version skew.41- Early cleanup can remove recovery.4243## Execution Checklist44451. Trace artifact identity, configuration, migration, compatibility, blast radius, and recovery ownership.462. Choose rollout, watch, containment, and rollback controls from reversibility and current policy.473. Verify mixed-version behavior, promotion provenance, stop signals, and recovery feasibility.484. **Analysis mode:** select rollout, watch, containment, and rollback controls.495. **Task mode:** produce artifact provenance, compatibility, and rollback metadata.506. **Review mode:** judge mixed-version, stop-signal, and recovery evidence.517. Stop when authority, artifact identity, or recovery proof remains implicit.5253## Stop / Escalation Conditions5455- Block when required immutable artifact identity is unproven.56- Block unverified target configuration, secrets, or environment equivalence.57- Require rollout approval, watch ownership, stop signals, and containment when risk triggers them.58- Require rollback rehearsal only when risk or policy demands it.59- Block migration or contract release lacking triggered mixed-version, ordering, coordination, reconciliation, rollback, or forward-repair proof.60- Block infrastructure release when desired/effective change, blast radius, authority, drift or state handling, or containment cannot be inspected.61- Refuse write tools without permission, sandbox, preview, recovery, and redaction evidence.6263## Output Contract6465- **Analysis mode (`analysis-agent`):** release plan; authority boundary; rollout and rollback decisions.66- **Task mode (`task-agent`):** release artifact; provenance, compatibility, and rollback metadata.67- **Review mode (`review-agent`):** go/no-go verdict; readiness gaps; unreviewed deployment risk.6869## Targeted References7071| Path | Type | Load when | Do not load when | Required by | Required output |72|---|---|---|---|---|---|73| [checklist](references/checklist.md) | decision-checklist | bounded L2 release checks remain unresolved | mode-specific closure or extended proof is required | analysis-agent, task-agent, review-agent | checklist-result, residual-risk |74| [delivery output and gates](references/delivery-output-and-gates.md) | targeted | L3-L5 release closure needs extended risk gates | compact L1-L2 release proof is sufficient | analysis-agent, task-agent, review-agent | gate-decision, residual-risk |75| [index](references/index.md) | index | competing release references require conflict or output-fragment selection | one task-named release reference resolves selection | analysis-agent, task-agent, review-agent | reference-selection |76| [release evidence](references/release-evidence-patterns.md) | evidence-pattern | release claims depend on runtime or artifact proof | no release claim needs fresh runtime evidence | analysis-agent, task-agent, review-agent | evidence-record, proof-limit, residual-risk |
machenjie