Choose what happens next at a workflow decision

<!-- sflow-copilot-pause -->

Reuse a verified pause-aware entry supplied in this invocation before any mutation or selection change; otherwise first run singularity-flow session current --for-agent --json once; pause precedes Git. If paused: native Copilot, only /sf-pause off, never resume implicitly. Reuse binding/personalization.replyName; fresh operation checks/consent remain required.

<!-- sflow-output-contract: governed-review -->

Output contract: Present artifact references, hashes, identity warnings, and the exact confirmation before recording any decision. Summarize artifacts with verified Markdown-preview links; never paste document bodies, code or diffs into Copilot. For code, list touched paths and observed added/deleted/created line counts; unavailable stays unavailable. Use existing evidence, not extra calls solely for display. Summaries, links and opened previews are not human review or approval. For every next action, honor returned commandGuidance in order: Copilot, Shell, VS Code (model-free) using modelFreeCommand; if absent, say "not available for this action". Missing Copilot: "no verified equivalent". Never invent routes. Run a check only when a returned field or command asks for it; never decide from a step's name, kind or route that it applies. A refusal saying a check does not apply to this step is not a blocker: skip it and continue. Act on a result's act first: its state, do routes and any ask. A bounded result names agentOutput.read ranges and omittedFields text files; read all in one response with your file tool. If your host saved a result to a file you cannot open, run that command again with --for-agent --json; never read the host's copy with shell tools.

<!-- sflow-execution-boundary -->

Boundary: reuse entry ready/workId, cwd=repositoryPath; CLI/workItemRoot paths only; never $HOME. No duplicate lookup. Use exact CLI-returned paths. repositoryPath is command cwd, not a prefix for absolute file paths. Never rebase paths or strip .git/worktrees. Issued private/external files are exact exceptions to repository confinement, not wider read/write permission; writes still need authorization. Resolve relative paths only against their declared base. On failure, preserve the path and error; never search or recreate it.

<!-- sflow-turn-boundary: decision-only -->

Turn boundary — decision-only: singularity-flow decision choose|applicability|scope|completeness|plan, singularity-flow upstream-issue decide and singularity-flow decisions record (issue options, after an explicit yes) are the only permitted mutations. Never edit files, run tests or builds, raw git, submit, approve, reject or next. The person types every choice and reason; never supply, infer or default one. List answers come from ask_user; the person accepts or replaces each default. Any refusal ends this turn.

1. Run singularity-flow decisions <WORK-ID> --for-agent --json. Non-empty data.items: do 11-14, then stop; use the linked issue/owner, never duplicate it. Else run singularity-flow decision show <WORK-ID> --json. Empty routing decisions is not approval readiness. Check pending (2), unsatisfied applicability (6), unlisted upstreamIssues (10), routed scope/completeness/plan (7-9). Only if clear, say nothing is waiting, show ahead, stop.
2. Show pending.label, why it waits (reason, round, maxRounds), who decides (by), values, and every option: id, label, toLabel, reach, skipLabels. With anyStep, a step ID or end may be named instead.
3. Ask for the option ID, or a step, and a reason.
4. Run singularity-flow decision choose <WORK-ID> --fetch --option <ID> --reason "<REASON>" --expected <pending.key>, or --to <STEP> in place of --option.
5. Report the commit, route, skipped steps, authority group and next phase; show Next in Copilot: /sf-next; stop.
6. For each applicability entry not satisfied, show responsibility, declaredReason and who decides (authority); ask why it does not apply. Run singularity-flow decision applicability <WORK-ID> --fetch --responsibility <RESPONSIBILITY> --reason "<REASON>"; report the commit; stop.
7. Scope statements: run singularity-flow evidence scope <WORK-ID> --json. For each unresolved item show id, text and sourceId; ask for a disposition (clauses for included or existing) and a reason. Run singularity-flow decision scope <WORK-ID> --fetch --item <ID> --as <DISPOSITION> [--clause <IDS>] --reason "<REASON>".
8. Completeness review, only when structurallyComplete: show inventorySha256; ask for each article (completeness, ambiguity, consistency, verifiability, boundary-conditions, non-functional) as satisfied, exception or not-applicable. Run singularity-flow decision completeness <WORK-ID> --fetch --confirm <inventorySha256> --article <ID>=<DECISION>... --reason "<REASON>". A review, not proof of correct scope.
9. Unplanned path: ask for its row, or its supporting class and why. Run singularity-flow decision plan <WORK-ID> --fetch --add-location <CLAUSE>=<PATH> --reason "<REASON>".
10. Upstream issues: show each id, phase, about. Outcomes: plan-correction (--link PRC-NNN), amend (--link FAM-NNN, approved), return (opens its change request), accept-risk (--expires YYYY-MM-DD) or dismiss. Run singularity-flow upstream-issue decide <UPI-ID> --work-id <WORK-ID> --outcome <OUTCOME> --reason "<REASON>".
11. Lead with data.headline; show each item's title and why (IDs and flags stay in details/effects). Not yours (decidedBy.allowed false): say decidedBy.words, give decidedBy.share, skip it.
12. ask_user per other item: option labels (recommended first) plus Not now, then each needs value with its default. action is theirs to do: give action.command; never record it.
13. Preview: singularity-flow decisions <WORK-ID> --choose <ID>=<OPTION>... --for-agent --json, adding --reason|--expires|--cause <ID>=<VALUE> only for changed values; show data.preview.summary and refused words/share.
14. data.preview.record.recordInChat: after an explicit yes run data.preview.record.command once, exactly; report it. Otherwise give the person that command for their own terminal (reused runs and test-run risks are theirs to record). Never edit a digest; DECISIONS_PACKET_CHANGED: show the fresh list, restart at 11.