[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75FKEPHBNW2ZT8GRWBCZ2X9":3,"skill-same:01M75FKEPHBNW2ZT8GRWBCZ2X9":43},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":13,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":42},"01M75FKEPHBNW2ZT8GRWBCZ2X9","ship",null,"Implement approved SDD tasks through the quality gate to Done — the second of the two-command workflow. Run after \u002Fsdd-bootstrap:bootstrap and human task approval.","aharada54914\u002Fsdd-forge","aharada54914","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F40755043?v=4","01M75FFC4V7WXZF0PK9RYWHNGF","01M75FFC5YJJ11B7MDCRVWB9Z6",0,"MIT","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fship\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: ship\ndescription: Implement approved SDD tasks through the quality gate to Done — the second of the two-command workflow. Run after \u002Fsdd-bootstrap:bootstrap and human task approval.\ndisable-model-invocation: true\n---\n\n# SDD Ship\n\nOrchestrate approved tasks from implementation through quality verification to Done.\nThis skill does not write specifications; use `\u002Fsdd-bootstrap:bootstrap` for that.\n\n## Requested status dashboard\n\nWhen a dashboard is requested, the initiating parent follows\n`\u003Cinstalled-root>\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fstatus\u002FSKILL.md` before delegating\npresentation work. Bind ChatGPT to Sites or Claude to Artifacts using the\nparent service, retain the selected presentation identity, and never change it\nbecause a child uses another service\u002Fmodel. Read-only display never advances a\nworkflow gate. Missing presentation capability is reported independently of the\nalready-authorized development work. On entry\u002Fresume of a requested dashboard,\nrun the status skill's idempotent local runtime start for this project\u002Fenvironment\nbefore checking the retained parent presentation. No external permanent host\nis required.\n\n## Managed allocation boundary\n\nAffected policy roles: `implementer`.\n\nApply this boundary only when the caller explicitly selects managed routing\nwith a policy request or model controls, or this allocation continues that\nmanaged workflow. With no managed input, retain the legacy invocation and model\nhints below. Once managed is selected, do not switch that workflow to legacy\nbecause selection, validation or launch fails. No new environment toggle is\nrequired, and this boundary does not grant delegation or execution permission.\n\nFor a managed allocation, apply in the allocating parent before an unbound\nforked Skill\u002FAgent, native role, or fresh context receives work, and before any\nidentity reservation. This\nboundary takes precedence over later native\u002Fdefault launch guidance for these\nroles. It does not add an availability or authority gate.\n\nKeep an already admitted managed entry with its validated selection\u002Fcontrols and\nmatching role, host, run\u002Fsession and workload. Keep a concretely known matching\nbatch startup and fixed generated Codex configuration under their original\ncontracts. An already bound worker continues its original procedure: do not\nselect again, allocate again, or run this refusal inside that same worker.\nA bound parent does not bind a new child, role or session: each new allocation\nneeds its own admitted selection\u002Fcarrier binding or takes the refusal below.\n\nWhen this operation has no admitted binding of the selected model and effort to\nits actual native carrier, substitute the next affected policy role and actual\nhost (`claude-code`, `codex-cli`, `copilot-cli`, or `gemini-cli`) and execute:\n\n```bash\npython3 -I -B \"\u003Cinstalled-root>\"\u002Fplugins\u002Fsdd-implementation\u002Fscripts\u002Fmanaged-model-entry.py unsupported-entry \\\n  --role \"\u003Cpolicy-role>\" --host \"\u003Chost>\"\n```\n\nExit 2 with `model-policy-error\u002Fv1` \u002F `model-unavailable` stops that allocation\nbefore work or reservation. Do not fall through to an old model hint, alias,\nunbound native default or another parent session; do not count allocation alone\nas selection proof. This refusal does not assert host incapability and never\nauthorizes same-session reload. Original host-incapable-only implementation\nfallback evidence and all review isolation requirements remain unchanged.\n\nApply in the Lite\u002FFull implementation parent before the relevant single-task or batch allocation. Keep the bound Lite Claude dispatch and the downstream known-matching batch route. Later quality\u002Fcross-model handoffs retain their own boundaries and existing consent.\n\n\n## Invocation\n\nClaude Code:\n\n```txt\n\u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md#T-001\n\u002Fsdd-ship:ship\n\u002Fsdd-ship:ship --lite specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --full specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --verify specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --retro specs\u002F\u003Cfeature>\u002Ftasks.md\n```\n\nCodex:\n\n```txt\nUse the ship skill for specs\u002F\u003Cfeature>\u002Ftasks.md\nUse the ship skill (no argument — context-aware selection)\n```\n\n### Flags\n\n| Flag | Effect |\n|---|---|\n| `--lite` | Force lite track (lite-gate) regardless of AGENTS.md spec_profile |\n| `--full` | Force full track (quality-gate); verifies acceptance-tests.md + traceability.md exist |\n| `--verify` | Before quality-gate, run cross-model-verify for tasks that have `Cross-Model: enabled` |\n| `--retro` | Run workflow-retrospective after all tasks reach Done |\n\n**`--fix RT-NNN` is NOT a sdd-ship flag.** Apply a review ticket directly:\n`\u002Fsdd-quality-loop:fix-by-review-ticket docs\u002Freview-tickets\u002FRT-NNN.yml`\n\n## Preconditions\n\nBefore doing any work, verify that `AGENTS.md` exists at the repository root and\nthat `scripts\u002Fcheck-sdd-structure.sh` (or `.ps1`) reports no `missing:` items.\nIf either check fails, stop immediately and direct the user to run\n`\u002Fsdd-bootstrap:bootstrap adopt`. Do not proceed without it.\n\nDo **not** invoke `sdd-sudo`, create `SDD_SUDO`, or modify the `SDD_SUDO` file.\nSudo management is always a separate human action.\n\n## MCP Preflight (Advisory)\n\nAttempt `get_next_sdd_command` with no arguments; if it is unavailable or\nthe call fails, continue with the file-based flow below.\n\n`get_next_sdd_command` is read-only: it only reads repository state and\nreturns a suggestion, never writing or mutating anything. Its answer is\nadvisory — it does not decide the target, the track, or any step that\nfollows. `## Step 1 — Target Selection` below remains the sole decision\nprocedure, and nothing this step reads changes what `## Step 1` concludes.\n\nThis step is unconditional: apply it the same way regardless of invocation\nform (a `specs\u002F\u003Cfeature>\u002Ftasks.md` path argument, an optional `#T-NNN` task\nselector, or the zero-argument auto-select form) and regardless of track\n(full \u002F lite) — do not skip or gate it by invocation form or track. Do not\nsurface the unavailability or the failure to the user as a run failure.\n\nIf the probe's suggestion and the conclusion `## Step 1 — Target Selection`\nreaches disagree, state in the output both that a disagreement occurred and\nwhich source was acted on. Always act on the `## Step 1` conclusion in that\ncase, never the probe's.\n\n## Step 1 — Target Selection\n\nAn explicitly requested investigation, performance study, isolated prototype or\nhandoff may return to bootstrap's optional purpose playbook. A decision packet\ncan display pinned evidence; recheck its digest and inputs on resume. The packet\ndoes not replace this skill's target, track or approval checks and never sets Done.\n\n### With a path argument\n\nUse the given `specs\u002F\u003Cfeature>\u002Ftasks.md` (with optional `#T-NNN` task selector).\nIf `#T-NNN` is given, only process that single task.\n\n### Zero-argument (no path given)\n\n1. Read `AGENTS.md` at the repository root. Look for an `## Active Spec Directories`\n   section listing feature paths.\n2. For each listed path, check `specs\u002F\u003Cfeature>\u002Ftasks.md` for at least one task\n   with `Approval: Approved` and `Status: Planned` or `Status: In Progress`.\n3. **Exactly one match**: print `[sdd-ship AUTO-SELECT] specs\u002F\u003Cfeature>\u002Ftasks.md`\n   and proceed without asking the user.\n4. **Zero matches**: print \"No active features with Approval: Approved tasks found.\n   Run \u002Fsdd-bootstrap:bootstrap first.\" and stop.\n5. **Multiple matches**: list all candidates with task counts and print\n   \"Multiple active features found. Re-run: \u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md\"\n   and stop without touching any feature.\n\n## Step 2 — Track Detection\n\nThis is `sdd-ship`'s Capability-Mode-relevant entry point. Run the\nhook-activation handshake HERE, as the first Capability-Mode-relevant action,\nbefore trusting any Project Context content:\n\n\u003C!-- sdd:handshake-wiring v1 -->\n\n1. `check-hook-activation-handshake --emit-challenge` — returns a fresh\n   single-use nonce and the canary target `sdd\u002F.hook-canary-sentinel`. If a\n   sentinel from an earlier run is already present, perform and record one\n   cleanup attempt first; the new challenge proceeds either way.\n2. This agent session makes its **own** real tool-call attempt against that\n   canary target, using the per-runtime template the challenge carries, and\n   records the raw result verbatim. The tool never performs the probe write\n   itself.\n3. `check-hook-activation-handshake --verify-response --nonce \u003Cnonce>\n   --recorded-result \u003Cpath> --runtime \u003Cclaude-code|codex-cli|copilot-cli>`.\n4. `HOOK_ACTIVE` — continue to track resolution. Any other outcome stops\n   **Capability Mode only**, and which projects that stops is decided by the\n   gate table below, not by this step alone: stop with\n   `CAPABILITY_RUNTIME_UNAVAILABLE` if and ONLY if this project's Project\n   Context is physically present and valid (gate G2); if it is physically\n   absent, the project never entered Capability Mode, so continue on the\n   compatibility fallback as `DISABLED_LEGACY` (gates G3\u002FG4). From a\n   valid-Context project, never fall back to legacy behaviour silently.\n\n\nFor Codex, use the challenge's `recorded_result_template` and\n`capture_instructions` to save the actual denial with native `apply_patch`.\nShell commands embedding the protected canary path may themselves be denied.\n\n\u003C!-- \u002Fsdd:handshake-wiring -->\n\n**When the presence\u002Fvalidity probe runs.** The handshake itself never reads\nthe Project Context, so step 4 records an outcome without yet knowing which\ngate row applies. The physical-presence-and-validity probe that decides G1\u002FG2\nfrom G3\u002FG4 is the FIRST step of \"Track resolution\" below, and it runs THERE —\nafter the handshake, before any track is chosen. This section precedes track\nresolution because it is normative about the handshake's meaning; it is not an\ninstruction to probe earlier.\n\n### What a non-`HOOK_ACTIVE` handshake does and does not stop\n\nStep 4 above stops **Capability Mode**, not every invocation of this skill.\nWhich of the two it is depends on whether this project ever entered Capability\nMode, and the table below is normative in BOTH directions:\n\n\u003C!-- sdd:capability-gate-scope v1 -->\n\n| Gate | Project Context | Handshake | Resolution |\n|---|---|---|---|\n| G1 | physically present and valid | `HOOK_ACTIVE` | `CAPABILITY_MODE` |\n| G2 | physically present and valid | not `HOOK_ACTIVE` | `CAPABILITY_RUNTIME_UNAVAILABLE` |\n| G3 | physically absent | `HOOK_ACTIVE` | `DISABLED_LEGACY` |\n| G4 | physically absent | not `HOOK_ACTIVE` | `DISABLED_LEGACY` |\n\n\u003C!-- \u002Fsdd:capability-gate-scope -->\n\n- **G2 — stop.** A project that HAS a valid Project Context has entered\n  Capability Mode, so a handshake that could not observe a genuine,\n  nonce-matched denial MUST stop with `CAPABILITY_RUNTIME_UNAVAILABLE`, and\n  MUST NOT degrade to the compatibility-fallback path below. That degradation\n  is precisely the silent downgrade ADR-0023 exists to close: `design.md:1112`\n  requires the calling skill to stop **Capability Mode**, never to silently\n  fall back to legacy.\n- **G4 — continue.** A project that has NO Project Context never entered\n  Capability Mode. It sits in ADR-0016's `disabled-legacy`, which is \"a\n  normal, expected condition for a project with no Project Context, not an\n  error\" (`requirements.md:1821-1827`), and which\n  `CAPABILITY_RUNTIME_UNAVAILABLE` is \"never conflated with\"\n  (`design.md:1734`). The handshake gates `HOOK_ACTIVE`-gated behaviour only\n  (`requirements.md:1078`), so a non-`HOOK_ACTIVE` result does not stop such a\n  project: it continues on the compatibility fallback below.\n\nThe prohibited transition is Capability Mode → legacy. Legacy → legacy is not\na downgrade, and G4 must never be reported as `CAPABILITY_RUNTIME_UNAVAILABLE`\nmerely because G2 is.\n\n### Track resolution\n\nResolve the track next. **Physical presence is decided FIRST and alone;\nREQ-005 approval validation runs SECOND, only once the file is known to\nexist.** A `sdd\u002Fproject-context.yaml` that exists but fails\n`validate-approval-sidecar` is **not** the same as one that is absent —\ntreating the two alike is the fail-open ADR-0023 closes.\n\n\u003C!-- sdd:track-selection-contract v1 -->\n\n| Case | Project Context | Flag | Resolution |\n|---|---|---|---|\n| C1 | physically absent | `--full`, `--lite`, or none | `COMPATIBILITY_FALLBACK` |\n| C2 | physically present, REQ-005 validation fails | `--full`, `--lite`, or none | `PROJECT_CONTEXT_INVALID` |\n| C3 | physically present and valid, `spec_profile: lite` | `--full` | `PROMOTE_FULL` |\n| C4 | physically present and valid, `spec_profile: lite` | `--lite` | `NO_OP_LITE` |\n| C5 | physically present and valid, `spec_profile: full` | `--lite` | `ERROR_STOP` |\n| C6 | physically present and valid, `spec_profile: full` | `--full` | `NO_OP_FULL` |\n\n\u003C!-- \u002Fsdd:track-selection-contract -->\n\n- `COMPATIBILITY_FALLBACK` (C1 only) — apply the compatibility fallback below,\n  unchanged from the pre-ADR-0023 contract.\n- `PROJECT_CONTEXT_INVALID` (C2) — stop before any task is started and print\n  `[sdd-ship] PROJECT_CONTEXT_INVALID: sdd\u002Fproject-context.yaml is present but\n  its approval could not be verified.` Do not start implementation, do not\n  invoke any gate, do not change any `Status`, and do not fall through to C1's\n  fallback or to an implicit `full`\u002F`lite` selection.\n- `PROMOTE_FULL` (C3) — resolved track is `full`. Print\n  `[sdd-ship] Track: full (PROMOTE_FULL)`, then verify the full-track\n  artifacts as described below.\n- `NO_OP_FULL` (C6) — resolved track is `full`. Print\n  `[sdd-ship] Track: full (NO_OP_FULL)`, then verify the full-track artifacts\n  as described below.\n- `NO_OP_LITE` (C4) — resolved track is `lite`. Print\n  `[sdd-ship] Track: lite (NO_OP_LITE)` and run the risk-upgrade scan below\n  before starting any task.\n- `ERROR_STOP` (C5) — `--lite` against a declared `full` profile would loosen\n  the declared profile. Stop before any task is started and print\n  `[sdd-ship] ERROR_STOP: --lite cannot downgrade a Project Context declaring\n  spec_profile: full. Re-run without --lite.` It is never silently ignored and\n  never silently honoured.\n\n**Full-track artifact verification** (`PROMOTE_FULL`, `NO_OP_FULL`, and the\nfallback's `--full` route): verify that `specs\u002F\u003Cfeature>\u002Facceptance-tests.md`\nand `specs\u002F\u003Cfeature>\u002Ftraceability.md` both exist. If either is missing, stop\nand instruct the user to run full-track sdd-bootstrap.\n\n`PLUGIN-CONTRACTS.md`'s Track Detection section is the normative source for\nthis table.\n\n### Risk-upgrade scan\n\nKeep the selected task block and requirements complete, including exclusions and\napproval notes. `check-risk-upgrade` owns the bounded non-requirement grammar in\n`sdd-lite\u002Freferences\u002Frisk-upgrade-policy.md`; do not prefilter sections or reword\nrisks to pass. Positive requirements override nominal exclusions during review.\n\n\nA resolved `full` track is the only scan bypass. For every route that could\nresolve to **LITE** — `NO_OP_LITE`, or any remaining default, profile, or\n`--lite` route of the compatibility fallback — build a local UTF-8 input by\nconcatenating the selected task's complete `## T-NNN` block followed by that\nfeature's `requirements.md`. Both inputs are mandatory; do not retrieve remote\nissue text and do not use a partial task block.\n\nRun the platform-local checker before considering a lite branch:\n\n```txt\nplugins\u002Fsdd-lite\u002Fscripts\u002Fcheck-risk-upgrade.sh \u003Ctask-block-plus-requirements>\npowershell -NoProfile -ExecutionPolicy Bypass -File plugins\u002Fsdd-lite\u002Fscripts\u002Fcheck-risk-upgrade.ps1 -Path \u003Ctask-block-plus-requirements>\n```\n\n- Exit 10 with `full-required: \u003Cprimary-id>; triggers=\u003Cordered-ids>` forces\n  **FULL** track even when `--lite` was requested. Print the checker diagnostic\n  followed by `[sdd-ship] Track: full (risk-upgrade)` and verify the full-track\n  `acceptance-tests.md` and `traceability.md` artifacts before task start. If either is absent, stop before task start and print `[sdd-ship] Full-track artifacts unavailable. Run \u002Fsdd-bootstrap:bootstrap for the full track.`\n- Exit 2 with `risk-upgrade: input unavailable` is fail-closed: stop before\n  implementation or the lite gate, print that diagnostic, and direct the user\n  to restore the selected task block and requirements or run full bootstrap.\n- Exit 0 with `lite-eligible` allows the resolved lite track to proceed.\n\n### Compatibility fallback (no Project Context)\n\nReached only from case C1, and unchanged from the pre-ADR-0023 contract.\nExecute in this priority order (first match wins):\n\n1. `--full` flag present → **FULL** track. Print `[sdd-ship] Track: full (--full override)`.\n   Verify that `specs\u002F\u003Cfeature>\u002Facceptance-tests.md` and `specs\u002F\u003Cfeature>\u002Ftraceability.md`\n   both exist. If either is missing, stop and instruct the user to run full-track\n   sdd-bootstrap or drop `--full` and use `--lite`.\n2. `--lite` flag present → **LITE** track. Print `[sdd-ship] Track: lite (--lite override)`.\n3. Read `AGENTS.md`. If `spec_profile: lite` appears on any line → **LITE** track.\n   Print `[sdd-ship] Track: lite (spec_profile: lite in AGENTS.md)`.\n4. Default → **FULL** track. Print `[sdd-ship] Track: full (no lite profile detected)`.\n\nThe track message is always printed **before** any task is started.\nIf the wrong track is selected, the user can cancel (Ctrl-C) and re-run with\nthe correct `--lite` or `--full` flag.\n\n## Step 3 — Implementation Loop\n\n### Full track\n\nInvoke `\u002Fsdd-implementation:implement-tasks specs\u002F\u003Cfeature>\u002Ftasks.md`.\nThat skill handles dependency ordering, per-task state transitions\n(Planned → In Progress → Implementation Complete), and implementation reports.\n\nDo **not** re-implement the dependency resolution logic here; delegate entirely\nto implement-tasks.\n\n### Lite track\n\nInvoke `\u002Fsdd-implementation:implement-task specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` for\neach Approved + Planned\u002FIn-Progress task in tasks.md order. Lite track has no\ndependency graph; implement tasks in document order.\n\n#### Managed Claude dispatch (opt-in)\n\nFor explicitly selected managed routing on Claude Code, run the entry below from the parent orchestrator\nbefore invoking the implement-task Skill, which could otherwise start\nimplementation in this context. For each Approved + Planned\u002FIn-Progress task in\ntasks.md order, write `\u003Cprompt-file>` under `\u003Crepository-root>` naming the exact\n`specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` and that task's Scope, compute its raw SHA-256\nas `\u003Cprompt-sha256>`, use a fresh canonical UUID for `\u003Csession-id>` and a new\noutput directory, and supply the current scoped policy request and registry, the\nexplicit installed Claude executable and the existing verified configuration\nobservation unchanged. The entry loads the complete original implement-task\nSkill and its seven fixed supporting sources into one fresh single-task context,\nwhich executes the Skill once as its orchestrator and status writer under\nnormal native permissions. This parent only dispatches and receives: it does not\nimplement the task or repeat the child's status, report or approval handling. A\ntyped failure stops this handoff; do not retry through an unmanaged Skill or\nAgent call. On success continue the existing flow.\n\n```bash\ncd \"\u003Crepository-root>\" || exit 1\npython3 -B \"\u003Cinstalled-root>\u002Fplugins\u002Fsdd-implementation\u002Fscripts\u002Fmanaged-model-entry.py\" claude-entry \\\n  --request \"\u003Cpolicy-request>\" --registry \"\u003Cregistry>\" \\\n  --role implementer --host claude-code --run-id \"\u003Crun-id>\" \\\n  --session-id \"\u003Csession-id>\" --attempt \"\u003Cattempt>\" --output-dir \"\u003Coutput-dir>\" \\\n  --require-actual-id false --executable \"\u003Cclaude-executable>\" \\\n  --prompt-file \"\u003Cprompt-file>\" --prompt-sha256 \"\u003Cprompt-sha256>\"\n```\n\nWhen managed inputs are absent, use the legacy route above. A managed allocation\non another host needs its own admitted binding or the refusal in the allocation\nboundary; an unsupported managed route cannot retry through legacy.\n\n### Resume behaviour\n\nOn re-invocation after interruption, skip tasks already at `Implementation Complete`\nor `Done`. Resume from the first eligible task (Approved + Planned or In Progress).\n\n## Step 4 — Quality Gate Loop\n\nExecute after all targeted tasks reach `Implementation Complete`.\n\n### Full track\n\nFor each task at `Implementation Complete`, in tasks.md document order:\n\n1. **Cross-model verification runs before the gate.** A task whose tasks.md\n   entry carries `Risk: critical` or `Security-Sensitive: true` MUST have\n   cross-model verification run in this same ship invocation before it reaches\n   the quality gate, **regardless of whether `--verify` was passed**. For each\n   such task, invoke\n   `\u002Fsdd-quality-loop:cross-model-verify specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` before\n   the gate.\n   - **Human-gated waiver.** Skip the required cross-model run for such a task\n     only when its tasks.md entry records a `Cross-Model-Waiver:` that is VALID.\n     A waiver is valid only when the SAME task also carries a human\n     `Approval: Approved` audit mark naming a second distinct human approver\n     (the human-only mark the deterministic guard already prevents an agent from\n     writing, and the same distinct-approver rule already mandated for critical\n     tasks). A `Cross-Model-Waiver:` that lacks that co-located human-approval\n     context is treated as ABSENT (fail-closed) and cross-model verification\n     remains required. An agent cannot grant itself a waiver, because it cannot\n     produce the human approval mark the waiver depends on. Guard-level\n     enforcement of the waiver token itself is out of scope here (future\n     hardening); the waiver is inert without the human approval context.\n   - **Fail-closed stop.** If such a task has neither a completed cross-model\n     run in this invocation nor a valid waiver, STOP the ship flow before the\n     quality gate and print a diagnostic naming the task:\n     `[sdd-ship] Cross-model verification required for T-NNN (Risk: critical or\n     Security-Sensitive: true), but it neither ran this invocation nor carries a\n     valid Cross-Model-Waiver: (a valid waiver needs a co-located human\n     Approval: Approved mark naming a second distinct approver). Run cross-model\n     verification or record a valid waiver, then re-invoke.`\n\n   Independently of the requirement above, if the `--verify` flag was passed and\n   the task has `Cross-Model: enabled` in its\n   tasks.md entry: invoke `\u002Fsdd-quality-loop:cross-model-verify specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`\n   before the gate.\n   - If no task in the batch has `Cross-Model: enabled`, print a warning:\n     `[sdd-ship] --verify passed but no tasks have Cross-Model: enabled. Add\n     the field to task entries that require panelist verification.`\n   - On lite track, `--verify` is silently ignored with the same warning.\n2. Invoke `\u002Fsdd-quality-loop:quality-gate specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`.\n3. **PASS (Done)**: task transitions to Done. Continue to next task.\n4. **BLOCKED (review tickets created)**: stop immediately. Print:\n   `[sdd-ship] Quality gate blocked for T-NNN. Review tickets written to\n   docs\u002Freview-tickets\u002F. Address each ticket, then re-invoke:\n   \u002Fsdd-quality-loop:fix-by-review-ticket docs\u002Freview-tickets\u002FRT-NNN.yml`\n   Do not proceed to other tasks while any task is Blocked.\n5. Cycle limit (disk-based — survives re-invocation across sessions): before\n   invoking quality-gate, run the deterministic cycle-limit script for this task\n   instead of counting reports by hand:\n\n   ```txt\n   plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fcheck-quality-gate-cycle-limit.sh T-NNN \u003Cfeature>\n   ```\n\n   On a Windows host without `python3`\u002F`node`, run the PowerShell twin instead:\n\n   ```txt\n   pwsh plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fcheck-quality-gate-cycle-limit.ps1 T-NNN \u003Cfeature>\n   ```\n\n   The script counts THIS FEATURE's existing gate reports under\n   `reports\u002Fquality-gate\u002F` (word-boundary match on the task id so `T-001` does\n   not match `T-0010`, AND an anchored match on this feature's own `Feature:`\n   header line, so a different feature's report sharing the same bare task id\n   is never counted; an absent directory counts zero) and prints its decision:\n   - `continue` (exit 0): fewer than three reports — proceed to invoke\n     quality-gate for T-NNN.\n   - `Escalate-Human` (non-zero exit): three or more reports and the task is\n     still not Done — stop autonomous retries and surface the remaining\n     findings for a human decision. Do not automatically invoke quality-gate\n     or fix-by-review-ticket again. If a human has explicitly authorized the\n     applicable correction and further review, follow **Resuming After A Human\n     Decision** in `plugins\u002Fsdd-quality-loop\u002Fskills\u002Fquality-gate\u002FSKILL.md`:\n     record the authorizer, instruction source, feature\u002Ftask, correction scope\n     and remaining findings, then obtain fresh independent review and all\n     required gate evidence. A general wish to finish is insufficient. Preserve\n     the limit's non-zero exit and all report history; this is a documented\n     human-authorized continuation, not a successful cycle-limit check.\n\n   Because each quality-gate run writes a durable timestamped report under\n   `reports\u002Fquality-gate\u002F`, this limit is computed from disk and holds across\n   separate `\u002Fsdd-ship:ship` invocations and sessions (it does not reset per\n   invocation).\n\n### Lite track\n\nFor each task at `Implementation Complete`:\n- **Ineligibility check (before the lite gate).** If the task's tasks.md entry\n  carries `Risk: critical` or `Security-Sensitive: true`, it is INELIGIBLE for\n  the lite track, because the lite track has no cross-model step. The lite gate\n  rejects it: STOP and print a diagnostic naming the task and directing the\n  human to the full track, e.g.\n  `[sdd-ship] T-NNN is Risk: critical or Security-Sensitive: true and is\n  ineligible for the lite track (no cross-model step). Re-run on the full track:\n  \u002Fsdd-ship:ship --full specs\u002F\u003Cfeature>\u002Ftasks.md`\n  Do not admit such a task to the lite gate.\nInvoke `\u002Fsdd-lite:lite-gate specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`.\n- PASS: task is Done.\n- FAIL: stop, surface the report, and instruct the user to fix and re-run.\n\n## Field Definitions\n\nThese optional per-task fields live in a tasks.md task entry. They are additive;\nexisting consumers ignore unknown fields.\n\n- `Security-Sensitive:` — optional per-task boolean trigger. `true` marks the\n  task as requiring cross-model verification regardless of its `Risk:` tier (see\n  Step 4). The task author proposes it; the human confirms it at approval.\n  Absent or `false` means the field does not force cross-model verification (the\n  `Risk:` tier still may).\n- `Cross-Model-Waiver:` — optional per-task field recording an explicit decision\n  to skip the cross-model verification a task would otherwise require. Who may\n  set it: a human only. Its audit value is a short human-authored reason. It is\n  honored only when the SAME task also carries a human `Approval: Approved`\n  audit mark naming a second distinct human approver — the human-only approval\n  context that makes the waiver valid; without that context the waiver is\n  ignored and cross-model verification remains required (fail-closed). Because\n  an agent cannot produce the human approval mark, an agent-written waiver has\n  no effect.\n- Lite-track rule for critical\u002Fsecurity-sensitive tasks: a task with\n  `Risk: critical` or `Security-Sensitive: true` is INELIGIBLE for the lite\n  track, because the lite track has no cross-model step. The lite gate rejects\n  such a task with a diagnostic naming the task and directing the human to the\n  full track (see Step 4, Lite track), rather than admitting it to a track that\n  would silently drop the cross-model requirement.\n\n## Step 5 — Completion Check\n\nAfter all targeted tasks reach Done:\n\n- If `--retro` flag was passed, or if all tasks in the feature (not just the\n  targeted subset) have reached Done for the first time in this session:\n  invoke `\u002Fsdd-quality-loop:workflow-retrospective specs\u002F\u003Cfeature>`.\n- Print a summary: number of tasks implemented, number Done, any remaining\n  non-Done tasks (with their current Status).\n\n## State Machine Summary\n\n```\nIDLE\n→ TARGET_SELECTED    (path argument or zero-arg auto-select)\n→ TRACK_DETECTED     (print [sdd-ship] Track: ... message)\n→ IMPLEMENTING       (implement-tasks for full \u002F implement-task loop for lite)\n→ QUALITY_GATE_LOOP  (quality-gate or lite-gate per task)\n   ├── task Done → continue loop\n   ├── task Blocked → STOP (surface review tickets)\n   └── ≥3 disk gate reports for task and not Done → STOP (Escalate-Human)\n       └── explicit applicable human decision → documented scoped continuation\n→ COMPLETION_CHECK   (all targeted tasks Done?)\n   └── yes → [--retro or all-Done] RETROSPECTIVE → DONE\n```\n\n## Context Compaction\n\nShip sessions span many tasks. All loop state is persisted on disk (tasks.md\nStatus fields, `reports\u002Fquality-gate\u002F`, review tickets), and both the Resume\nbehaviour (Step 3) and the disk-based cycle limit (Step 4) are designed to\nsurvive re-invocation — compaction at a task boundary loses nothing.\n\nCompact (or accept auto-compaction) only at task boundaries:\n\n- after a task reaches Done (quality gate PASS)\n- after a BLOCKED stop, before starting fix-by-review-ticket\n- between Step 3 and Step 4, when all targeted tasks are at\n  Implementation Complete\n\nDo not compact while a task is In Progress or a quality-gate cycle is\nrunning: the implementing agent's working context (file paths, partial\nchanges, verification state) is not yet reflected on disk. After compacting,\nre-invoke `\u002Fsdd-ship:ship` and rely on Resume behaviour to continue from the\nfirst eligible task.\n\n## Security Boundaries\n\n- **Never** invoke `sdd-sudo` or create\u002Fmodify the `SDD_SUDO` file.\n- **Never** set `Approval: Approved` on any task. The approval guard will block\n  this; attempting it causes confusing error messages.\n- **Never** modify gate scripts (`scripts\u002F*.sh`, `scripts\u002F*.ps1`) or hook files.\n- **Never** push to remote, create pull requests, or merge branches unless the\n  user explicitly requests it in this session.\n- **Never** modify files under `plugins\u002Fsdd-quality-loop\u002Fhooks\u002F` or\n  `plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fsdd-hook-guard.*`.\n- Do not exceed the scope of the targeted tasks. Do not make changes to files\n  unrelated to the approved tasks.\n\n## Handoff\n\nAfter sdd-ship completes, report:\n\n- Tasks implemented and their final status (Done \u002F Blocked)\n- Any review tickets created (with file paths)\n- Next action for the user:\n  - If all Done: \"Feature complete. Consider re-running \u002Fsdd-ship:ship --retro\n    specs\u002F\u003Cfeature>\u002Ftasks.md to capture improvements.\"\n  - If Blocked: \"Address review tickets, apply fixes with\n    \u002Fsdd-quality-loop:fix-by-review-ticket, then re-run \u002Fsdd-ship:ship.\"\n","# SDD Ship\n\nOrchestrate approved tasks from implementation through quality verification to Done.\nThis skill does not write specifications; use `\u002Fsdd-bootstrap:bootstrap` for that.\n\n## Requested status dashboard\n\nWhen a dashboard is requested, the initiating parent follows\n`\u003Cinstalled-root>\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fstatus\u002FSKILL.md` before delegating\npresentation work. Bind ChatGPT to Sites or Claude to Artifacts using the\nparent service, retain the selected presentation identity, and never change it\nbecause a child uses another service\u002Fmodel. Read-only display never advances a\nworkflow gate. Missing presentation capability is reported independently of the\nalready-authorized development work. On entry\u002Fresume of a requested dashboard,\nrun the status skill's idempotent local runtime start for this project\u002Fenvironment\nbefore checking the retained parent presentation. No external permanent host\nis required.\n\n## Managed allocation boundary\n\nAffected policy roles: `implementer`.\n\nApply this boundary only when the caller explicitly selects managed routing\nwith a policy request or model controls, or this allocation continues that\nmanaged workflow. With no managed input, retain the legacy invocation and model\nhints below. Once managed is selected, do not switch that workflow to legacy\nbecause selection, validation or launch fails. No new environment toggle is\nrequired, and this boundary does not grant delegation or execution permission.\n\nFor a managed allocation, apply in the allocating parent before an unbound\nforked Skill\u002FAgent, native role, or fresh context receives work, and before any\nidentity reservation. This\nboundary takes precedence over later native\u002Fdefault launch guidance for these\nroles. It does not add an availability or authority gate.\n\nKeep an already admitted managed entry with its validated selection\u002Fcontrols and\nmatching role, host, run\u002Fsession and workload. Keep a concretely known matching\nbatch startup and fixed generated Codex configuration under their original\ncontracts. An already bound worker continues its original procedure: do not\nselect again, allocate again, or run this refusal inside that same worker.\nA bound parent does not bind a new child, role or session: each new allocation\nneeds its own admitted selection\u002Fcarrier binding or takes the refusal below.\n\nWhen this operation has no admitted binding of the selected model and effort to\nits actual native carrier, substitute the next affected policy role and actual\nhost (`claude-code`, `codex-cli`, `copilot-cli`, or `gemini-cli`) and execute:\n\n```bash\npython3 -I -B \"\u003Cinstalled-root>\"\u002Fplugins\u002Fsdd-implementation\u002Fscripts\u002Fmanaged-model-entry.py unsupported-entry \\\n  --role \"\u003Cpolicy-role>\" --host \"\u003Chost>\"\n```\n\nExit 2 with `model-policy-error\u002Fv1` \u002F `model-unavailable` stops that allocation\nbefore work or reservation. Do not fall through to an old model hint, alias,\nunbound native default or another parent session; do not count allocation alone\nas selection proof. This refusal does not assert host incapability and never\nauthorizes same-session reload. Original host-incapable-only implementation\nfallback evidence and all review isolation requirements remain unchanged.\n\nApply in the Lite\u002FFull implementation parent before the relevant single-task or batch allocation. Keep the bound Lite Claude dispatch and the downstream known-matching batch route. Later quality\u002Fcross-model handoffs retain their own boundaries and existing consent.\n\n\n## Invocation\n\nClaude Code:\n\n```txt\n\u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md#T-001\n\u002Fsdd-ship:ship\n\u002Fsdd-ship:ship --lite specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --full specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --verify specs\u002F\u003Cfeature>\u002Ftasks.md\n\u002Fsdd-ship:ship --retro specs\u002F\u003Cfeature>\u002Ftasks.md\n```\n\nCodex:\n\n```txt\nUse the ship skill for specs\u002F\u003Cfeature>\u002Ftasks.md\nUse the ship skill (no argument — context-aware selection)\n```\n\n### Flags\n\n| Flag | Effect |\n|---|---|\n| `--lite` | Force lite track (lite-gate) regardless of AGENTS.md spec_profile |\n| `--full` | Force full track (quality-gate); verifies acceptance-tests.md + traceability.md exist |\n| `--verify` | Before quality-gate, run cross-model-verify for tasks that have `Cross-Model: enabled` |\n| `--retro` | Run workflow-retrospective after all tasks reach Done |\n\n**`--fix RT-NNN` is NOT a sdd-ship flag.** Apply a review ticket directly:\n`\u002Fsdd-quality-loop:fix-by-review-ticket docs\u002Freview-tickets\u002FRT-NNN.yml`\n\n## Preconditions\n\nBefore doing any work, verify that `AGENTS.md` exists at the repository root and\nthat `scripts\u002Fcheck-sdd-structure.sh` (or `.ps1`) reports no `missing:` items.\nIf either check fails, stop immediately and direct the user to run\n`\u002Fsdd-bootstrap:bootstrap adopt`. Do not proceed without it.\n\nDo **not** invoke `sdd-sudo`, create `SDD_SUDO`, or modify the `SDD_SUDO` file.\nSudo management is always a separate human action.\n\n## MCP Preflight (Advisory)\n\nAttempt `get_next_sdd_command` with no arguments; if it is unavailable or\nthe call fails, continue with the file-based flow below.\n\n`get_next_sdd_command` is read-only: it only reads repository state and\nreturns a suggestion, never writing or mutating anything. Its answer is\nadvisory — it does not decide the target, the track, or any step that\nfollows. `## Step 1 — Target Selection` below remains the sole decision\nprocedure, and nothing this step reads changes what `## Step 1` concludes.\n\nThis step is unconditional: apply it the same way regardless of invocation\nform (a `specs\u002F\u003Cfeature>\u002Ftasks.md` path argument, an optional `#T-NNN` task\nselector, or the zero-argument auto-select form) and regardless of track\n(full \u002F lite) — do not skip or gate it by invocation form or track. Do not\nsurface the unavailability or the failure to the user as a run failure.\n\nIf the probe's suggestion and the conclusion `## Step 1 — Target Selection`\nreaches disagree, state in the output both that a disagreement occurred and\nwhich source was acted on. Always act on the `## Step 1` conclusion in that\ncase, never the probe's.\n\n## Step 1 — Target Selection\n\nAn explicitly requested investigation, performance study, isolated prototype or\nhandoff may return to bootstrap's optional purpose playbook. A decision packet\ncan display pinned evidence; recheck its digest and inputs on resume. The packet\ndoes not replace this skill's target, track or approval checks and never sets Done.\n\n### With a path argument\n\nUse the given `specs\u002F\u003Cfeature>\u002Ftasks.md` (with optional `#T-NNN` task selector).\nIf `#T-NNN` is given, only process that single task.\n\n### Zero-argument (no path given)\n\n1. Read `AGENTS.md` at the repository root. Look for an `## Active Spec Directories`\n   section listing feature paths.\n2. For each listed path, check `specs\u002F\u003Cfeature>\u002Ftasks.md` for at least one task\n   with `Approval: Approved` and `Status: Planned` or `Status: In Progress`.\n3. **Exactly one match**: print `[sdd-ship AUTO-SELECT] specs\u002F\u003Cfeature>\u002Ftasks.md`\n   and proceed without asking the user.\n4. **Zero matches**: print \"No active features with Approval: Approved tasks found.\n   Run \u002Fsdd-bootstrap:bootstrap first.\" and stop.\n5. **Multiple matches**: list all candidates with task counts and print\n   \"Multiple active features found. Re-run: \u002Fsdd-ship:ship specs\u002F\u003Cfeature>\u002Ftasks.md\"\n   and stop without touching any feature.\n\n## Step 2 — Track Detection\n\nThis is `sdd-ship`'s Capability-Mode-relevant entry point. Run the\nhook-activation handshake HERE, as the first Capability-Mode-relevant action,\nbefore trusting any Project Context content:\n\n\u003C!-- sdd:handshake-wiring v1 -->\n\n1. `check-hook-activation-handshake --emit-challenge` — returns a fresh\n   single-use nonce and the canary target `sdd\u002F.hook-canary-sentinel`. If a\n   sentinel from an earlier run is already present, perform and record one\n   cleanup attempt first; the new challenge proceeds either way.\n2. This agent session makes its **own** real tool-call attempt against that\n   canary target, using the per-runtime template the challenge carries, and\n   records the raw result verbatim. The tool never performs the probe write\n   itself.\n3. `check-hook-activation-handshake --verify-response --nonce \u003Cnonce>\n   --recorded-result \u003Cpath> --runtime \u003Cclaude-code|codex-cli|copilot-cli>`.\n4. `HOOK_ACTIVE` — continue to track resolution. Any other outcome stops\n   **Capability Mode only**, and which projects that stops is decided by the\n   gate table below, not by this step alone: stop with\n   `CAPABILITY_RUNTIME_UNAVAILABLE` if and ONLY if this project's Project\n   Context is physically present and valid (gate G2); if it is physically\n   absent, the project never entered Capability Mode, so continue on the\n   compatibility fallback as `DISABLED_LEGACY` (gates G3\u002FG4). From a\n   valid-Context project, never fall back to legacy behaviour silently.\n\n\nFor Codex, use the challenge's `recorded_result_template` and\n`capture_instructions` to save the actual denial with native `apply_patch`.\nShell commands embedding the protected canary path may themselves be denied.\n\n\u003C!-- \u002Fsdd:handshake-wiring -->\n\n**When the presence\u002Fvalidity probe runs.** The handshake itself never reads\nthe Project Context, so step 4 records an outcome without yet knowing which\ngate row applies. The physical-presence-and-validity probe that decides G1\u002FG2\nfrom G3\u002FG4 is the FIRST step of \"Track resolution\" below, and it runs THERE —\nafter the handshake, before any track is chosen. This section precedes track\nresolution because it is normative about the handshake's meaning; it is not an\ninstruction to probe earlier.\n\n### What a non-`HOOK_ACTIVE` handshake does and does not stop\n\nStep 4 above stops **Capability Mode**, not every invocation of this skill.\nWhich of the two it is depends on whether this project ever entered Capability\nMode, and the table below is normative in BOTH directions:\n\n\u003C!-- sdd:capability-gate-scope v1 -->\n\n| Gate | Project Context | Handshake | Resolution |\n|---|---|---|---|\n| G1 | physically present and valid | `HOOK_ACTIVE` | `CAPABILITY_MODE` |\n| G2 | physically present and valid | not `HOOK_ACTIVE` | `CAPABILITY_RUNTIME_UNAVAILABLE` |\n| G3 | physically absent | `HOOK_ACTIVE` | `DISABLED_LEGACY` |\n| G4 | physically absent | not `HOOK_ACTIVE` | `DISABLED_LEGACY` |\n\n\u003C!-- \u002Fsdd:capability-gate-scope -->\n\n- **G2 — stop.** A project that HAS a valid Project Context has entered\n  Capability Mode, so a handshake that could not observe a genuine,\n  nonce-matched denial MUST stop with `CAPABILITY_RUNTIME_UNAVAILABLE`, and\n  MUST NOT degrade to the compatibility-fallback path below. That degradation\n  is precisely the silent downgrade ADR-0023 exists to close: `design.md:1112`\n  requires the calling skill to stop **Capability Mode**, never to silently\n  fall back to legacy.\n- **G4 — continue.** A project that has NO Project Context never entered\n  Capability Mode. It sits in ADR-0016's `disabled-legacy`, which is \"a\n  normal, expected condition for a project with no Project Context, not an\n  error\" (`requirements.md:1821-1827`), and which\n  `CAPABILITY_RUNTIME_UNAVAILABLE` is \"never conflated with\"\n  (`design.md:1734`). The handshake gates `HOOK_ACTIVE`-gated behaviour only\n  (`requirements.md:1078`), so a non-`HOOK_ACTIVE` result does not stop such a\n  project: it continues on the compatibility fallback below.\n\nThe prohibited transition is Capability Mode → legacy. Legacy → legacy is not\na downgrade, and G4 must never be reported as `CAPABILITY_RUNTIME_UNAVAILABLE`\nmerely because G2 is.\n\n### Track resolution\n\nResolve the track next. **Physical presence is decided FIRST and alone;\nREQ-005 approval validation runs SECOND, only once the file is known to\nexist.** A `sdd\u002Fproject-context.yaml` that exists but fails\n`validate-approval-sidecar` is **not** the same as one that is absent —\ntreating the two alike is the fail-open ADR-0023 closes.\n\n\u003C!-- sdd:track-selection-contract v1 -->\n\n| Case | Project Context | Flag | Resolution |\n|---|---|---|---|\n| C1 | physically absent | `--full`, `--lite`, or none | `COMPATIBILITY_FALLBACK` |\n| C2 | physically present, REQ-005 validation fails | `--full`, `--lite`, or none | `PROJECT_CONTEXT_INVALID` |\n| C3 | physically present and valid, `spec_profile: lite` | `--full` | `PROMOTE_FULL` |\n| C4 | physically present and valid, `spec_profile: lite` | `--lite` | `NO_OP_LITE` |\n| C5 | physically present and valid, `spec_profile: full` | `--lite` | `ERROR_STOP` |\n| C6 | physically present and valid, `spec_profile: full` | `--full` | `NO_OP_FULL` |\n\n\u003C!-- \u002Fsdd:track-selection-contract -->\n\n- `COMPATIBILITY_FALLBACK` (C1 only) — apply the compatibility fallback below,\n  unchanged from the pre-ADR-0023 contract.\n- `PROJECT_CONTEXT_INVALID` (C2) — stop before any task is started and print\n  `[sdd-ship] PROJECT_CONTEXT_INVALID: sdd\u002Fproject-context.yaml is present but\n  its approval could not be verified.` Do not start implementation, do not\n  invoke any gate, do not change any `Status`, and do not fall through to C1's\n  fallback or to an implicit `full`\u002F`lite` selection.\n- `PROMOTE_FULL` (C3) — resolved track is `full`. Print\n  `[sdd-ship] Track: full (PROMOTE_FULL)`, then verify the full-track\n  artifacts as described below.\n- `NO_OP_FULL` (C6) — resolved track is `full`. Print\n  `[sdd-ship] Track: full (NO_OP_FULL)`, then verify the full-track artifacts\n  as described below.\n- `NO_OP_LITE` (C4) — resolved track is `lite`. Print\n  `[sdd-ship] Track: lite (NO_OP_LITE)` and run the risk-upgrade scan below\n  before starting any task.\n- `ERROR_STOP` (C5) — `--lite` against a declared `full` profile would loosen\n  the declared profile. Stop before any task is started and print\n  `[sdd-ship] ERROR_STOP: --lite cannot downgrade a Project Context declaring\n  spec_profile: full. Re-run without --lite.` It is never silently ignored and\n  never silently honoured.\n\n**Full-track artifact verification** (`PROMOTE_FULL`, `NO_OP_FULL`, and the\nfallback's `--full` route): verify that `specs\u002F\u003Cfeature>\u002Facceptance-tests.md`\nand `specs\u002F\u003Cfeature>\u002Ftraceability.md` both exist. If either is missing, stop\nand instruct the user to run full-track sdd-bootstrap.\n\n`PLUGIN-CONTRACTS.md`'s Track Detection section is the normative source for\nthis table.\n\n### Risk-upgrade scan\n\nKeep the selected task block and requirements complete, including exclusions and\napproval notes. `check-risk-upgrade` owns the bounded non-requirement grammar in\n`sdd-lite\u002Freferences\u002Frisk-upgrade-policy.md`; do not prefilter sections or reword\nrisks to pass. Positive requirements override nominal exclusions during review.\n\n\nA resolved `full` track is the only scan bypass. For every route that could\nresolve to **LITE** — `NO_OP_LITE`, or any remaining default, profile, or\n`--lite` route of the compatibility fallback — build a local UTF-8 input by\nconcatenating the selected task's complete `## T-NNN` block followed by that\nfeature's `requirements.md`. Both inputs are mandatory; do not retrieve remote\nissue text and do not use a partial task block.\n\nRun the platform-local checker before considering a lite branch:\n\n```txt\nplugins\u002Fsdd-lite\u002Fscripts\u002Fcheck-risk-upgrade.sh \u003Ctask-block-plus-requirements>\npowershell -NoProfile -ExecutionPolicy Bypass -File plugins\u002Fsdd-lite\u002Fscripts\u002Fcheck-risk-upgrade.ps1 -Path \u003Ctask-block-plus-requirements>\n```\n\n- Exit 10 with `full-required: \u003Cprimary-id>; triggers=\u003Cordered-ids>` forces\n  **FULL** track even when `--lite` was requested. Print the checker diagnostic\n  followed by `[sdd-ship] Track: full (risk-upgrade)` and verify the full-track\n  `acceptance-tests.md` and `traceability.md` artifacts before task start. If either is absent, stop before task start and print `[sdd-ship] Full-track artifacts unavailable. Run \u002Fsdd-bootstrap:bootstrap for the full track.`\n- Exit 2 with `risk-upgrade: input unavailable` is fail-closed: stop before\n  implementation or the lite gate, print that diagnostic, and direct the user\n  to restore the selected task block and requirements or run full bootstrap.\n- Exit 0 with `lite-eligible` allows the resolved lite track to proceed.\n\n### Compatibility fallback (no Project Context)\n\nReached only from case C1, and unchanged from the pre-ADR-0023 contract.\nExecute in this priority order (first match wins):\n\n1. `--full` flag present → **FULL** track. Print `[sdd-ship] Track: full (--full override)`.\n   Verify that `specs\u002F\u003Cfeature>\u002Facceptance-tests.md` and `specs\u002F\u003Cfeature>\u002Ftraceability.md`\n   both exist. If either is missing, stop and instruct the user to run full-track\n   sdd-bootstrap or drop `--full` and use `--lite`.\n2. `--lite` flag present → **LITE** track. Print `[sdd-ship] Track: lite (--lite override)`.\n3. Read `AGENTS.md`. If `spec_profile: lite` appears on any line → **LITE** track.\n   Print `[sdd-ship] Track: lite (spec_profile: lite in AGENTS.md)`.\n4. Default → **FULL** track. Print `[sdd-ship] Track: full (no lite profile detected)`.\n\nThe track message is always printed **before** any task is started.\nIf the wrong track is selected, the user can cancel (Ctrl-C) and re-run with\nthe correct `--lite` or `--full` flag.\n\n## Step 3 — Implementation Loop\n\n### Full track\n\nInvoke `\u002Fsdd-implementation:implement-tasks specs\u002F\u003Cfeature>\u002Ftasks.md`.\nThat skill handles dependency ordering, per-task state transitions\n(Planned → In Progress → Implementation Complete), and implementation reports.\n\nDo **not** re-implement the dependency resolution logic here; delegate entirely\nto implement-tasks.\n\n### Lite track\n\nInvoke `\u002Fsdd-implementation:implement-task specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` for\neach Approved + Planned\u002FIn-Progress task in tasks.md order. Lite track has no\ndependency graph; implement tasks in document order.\n\n#### Managed Claude dispatch (opt-in)\n\nFor explicitly selected managed routing on Claude Code, run the entry below from the parent orchestrator\nbefore invoking the implement-task Skill, which could otherwise start\nimplementation in this context. For each Approved + Planned\u002FIn-Progress task in\ntasks.md order, write `\u003Cprompt-file>` under `\u003Crepository-root>` naming the exact\n`specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` and that task's Scope, compute its raw SHA-256\nas `\u003Cprompt-sha256>`, use a fresh canonical UUID for `\u003Csession-id>` and a new\noutput directory, and supply the current scoped policy request and registry, the\nexplicit installed Claude executable and the existing verified configuration\nobservation unchanged. The entry loads the complete original implement-task\nSkill and its seven fixed supporting sources into one fresh single-task context,\nwhich executes the Skill once as its orchestrator and status writer under\nnormal native permissions. This parent only dispatches and receives: it does not\nimplement the task or repeat the child's status, report or approval handling. A\ntyped failure stops this handoff; do not retry through an unmanaged Skill or\nAgent call. On success continue the existing flow.\n\n```bash\ncd \"\u003Crepository-root>\" || exit 1\npython3 -B \"\u003Cinstalled-root>\u002Fplugins\u002Fsdd-implementation\u002Fscripts\u002Fmanaged-model-entry.py\" claude-entry \\\n  --request \"\u003Cpolicy-request>\" --registry \"\u003Cregistry>\" \\\n  --role implementer --host claude-code --run-id \"\u003Crun-id>\" \\\n  --session-id \"\u003Csession-id>\" --attempt \"\u003Cattempt>\" --output-dir \"\u003Coutput-dir>\" \\\n  --require-actual-id false --executable \"\u003Cclaude-executable>\" \\\n  --prompt-file \"\u003Cprompt-file>\" --prompt-sha256 \"\u003Cprompt-sha256>\"\n```\n\nWhen managed inputs are absent, use the legacy route above. A managed allocation\non another host needs its own admitted binding or the refusal in the allocation\nboundary; an unsupported managed route cannot retry through legacy.\n\n### Resume behaviour\n\nOn re-invocation after interruption, skip tasks already at `Implementation Complete`\nor `Done`. Resume from the first eligible task (Approved + Planned or In Progress).\n\n## Step 4 — Quality Gate Loop\n\nExecute after all targeted tasks reach `Implementation Complete`.\n\n### Full track\n\nFor each task at `Implementation Complete`, in tasks.md document order:\n\n1. **Cross-model verification runs before the gate.** A task whose tasks.md\n   entry carries `Risk: critical` or `Security-Sensitive: true` MUST have\n   cross-model verification run in this same ship invocation before it reaches\n   the quality gate, **regardless of whether `--verify` was passed**. For each\n   such task, invoke\n   `\u002Fsdd-quality-loop:cross-model-verify specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN` before\n   the gate.\n   - **Human-gated waiver.** Skip the required cross-model run for such a task\n     only when its tasks.md entry records a `Cross-Model-Waiver:` that is VALID.\n     A waiver is valid only when the SAME task also carries a human\n     `Approval: Approved` audit mark naming a second distinct human approver\n     (the human-only mark the deterministic guard already prevents an agent from\n     writing, and the same distinct-approver rule already mandated for critical\n     tasks). A `Cross-Model-Waiver:` that lacks that co-located human-approval\n     context is treated as ABSENT (fail-closed) and cross-model verification\n     remains required. An agent cannot grant itself a waiver, because it cannot\n     produce the human approval mark the waiver depends on. Guard-level\n     enforcement of the waiver token itself is out of scope here (future\n     hardening); the waiver is inert without the human approval context.\n   - **Fail-closed stop.** If such a task has neither a completed cross-model\n     run in this invocation nor a valid waiver, STOP the ship flow before the\n     quality gate and print a diagnostic naming the task:\n     `[sdd-ship] Cross-model verification required for T-NNN (Risk: critical or\n     Security-Sensitive: true), but it neither ran this invocation nor carries a\n     valid Cross-Model-Waiver: (a valid waiver needs a co-located human\n     Approval: Approved mark naming a second distinct approver). Run cross-model\n     verification or record a valid waiver, then re-invoke.`\n\n   Independently of the requirement above, if the `--verify` flag was passed and\n   the task has `Cross-Model: enabled` in its\n   tasks.md entry: invoke `\u002Fsdd-quality-loop:cross-model-verify specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`\n   before the gate.\n   - If no task in the batch has `Cross-Model: enabled`, print a warning:\n     `[sdd-ship] --verify passed but no tasks have Cross-Model: enabled. Add\n     the field to task entries that require panelist verification.`\n   - On lite track, `--verify` is silently ignored with the same warning.\n2. Invoke `\u002Fsdd-quality-loop:quality-gate specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`.\n3. **PASS (Done)**: task transitions to Done. Continue to next task.\n4. **BLOCKED (review tickets created)**: stop immediately. Print:\n   `[sdd-ship] Quality gate blocked for T-NNN. Review tickets written to\n   docs\u002Freview-tickets\u002F. Address each ticket, then re-invoke:\n   \u002Fsdd-quality-loop:fix-by-review-ticket docs\u002Freview-tickets\u002FRT-NNN.yml`\n   Do not proceed to other tasks while any task is Blocked.\n5. Cycle limit (disk-based — survives re-invocation across sessions): before\n   invoking quality-gate, run the deterministic cycle-limit script for this task\n   instead of counting reports by hand:\n\n   ```txt\n   plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fcheck-quality-gate-cycle-limit.sh T-NNN \u003Cfeature>\n   ```\n\n   On a Windows host without `python3`\u002F`node`, run the PowerShell twin instead:\n\n   ```txt\n   pwsh plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fcheck-quality-gate-cycle-limit.ps1 T-NNN \u003Cfeature>\n   ```\n\n   The script counts THIS FEATURE's existing gate reports under\n   `reports\u002Fquality-gate\u002F` (word-boundary match on the task id so `T-001` does\n   not match `T-0010`, AND an anchored match on this feature's own `Feature:`\n   header line, so a different feature's report sharing the same bare task id\n   is never counted; an absent directory counts zero) and prints its decision:\n   - `continue` (exit 0): fewer than three reports — proceed to invoke\n     quality-gate for T-NNN.\n   - `Escalate-Human` (non-zero exit): three or more reports and the task is\n     still not Done — stop autonomous retries and surface the remaining\n     findings for a human decision. Do not automatically invoke quality-gate\n     or fix-by-review-ticket again. If a human has explicitly authorized the\n     applicable correction and further review, follow **Resuming After A Human\n     Decision** in `plugins\u002Fsdd-quality-loop\u002Fskills\u002Fquality-gate\u002FSKILL.md`:\n     record the authorizer, instruction source, feature\u002Ftask, correction scope\n     and remaining findings, then obtain fresh independent review and all\n     required gate evidence. A general wish to finish is insufficient. Preserve\n     the limit's non-zero exit and all report history; this is a documented\n     human-authorized continuation, not a successful cycle-limit check.\n\n   Because each quality-gate run writes a durable timestamped report under\n   `reports\u002Fquality-gate\u002F`, this limit is computed from disk and holds across\n   separate `\u002Fsdd-ship:ship` invocations and sessions (it does not reset per\n   invocation).\n\n### Lite track\n\nFor each task at `Implementation Complete`:\n- **Ineligibility check (before the lite gate).** If the task's tasks.md entry\n  carries `Risk: critical` or `Security-Sensitive: true`, it is INELIGIBLE for\n  the lite track, because the lite track has no cross-model step. The lite gate\n  rejects it: STOP and print a diagnostic naming the task and directing the\n  human to the full track, e.g.\n  `[sdd-ship] T-NNN is Risk: critical or Security-Sensitive: true and is\n  ineligible for the lite track (no cross-model step). Re-run on the full track:\n  \u002Fsdd-ship:ship --full specs\u002F\u003Cfeature>\u002Ftasks.md`\n  Do not admit such a task to the lite gate.\nInvoke `\u002Fsdd-lite:lite-gate specs\u002F\u003Cfeature>\u002Ftasks.md#T-NNN`.\n- PASS: task is Done.\n- FAIL: stop, surface the report, and instruct the user to fix and re-run.\n\n## Field Definitions\n\nThese optional per-task fields live in a tasks.md task entry. They are additive;\nexisting consumers ignore unknown fields.\n\n- `Security-Sensitive:` — optional per-task boolean trigger. `true` marks the\n  task as requiring cross-model verification regardless of its `Risk:` tier (see\n  Step 4). The task author proposes it; the human confirms it at approval.\n  Absent or `false` means the field does not force cross-model verification (the\n  `Risk:` tier still may).\n- `Cross-Model-Waiver:` — optional per-task field recording an explicit decision\n  to skip the cross-model verification a task would otherwise require. Who may\n  set it: a human only. Its audit value is a short human-authored reason. It is\n  honored only when the SAME task also carries a human `Approval: Approved`\n  audit mark naming a second distinct human approver — the human-only approval\n  context that makes the waiver valid; without that context the waiver is\n  ignored and cross-model verification remains required (fail-closed). Because\n  an agent cannot produce the human approval mark, an agent-written waiver has\n  no effect.\n- Lite-track rule for critical\u002Fsecurity-sensitive tasks: a task with\n  `Risk: critical` or `Security-Sensitive: true` is INELIGIBLE for the lite\n  track, because the lite track has no cross-model step. The lite gate rejects\n  such a task with a diagnostic naming the task and directing the human to the\n  full track (see Step 4, Lite track), rather than admitting it to a track that\n  would silently drop the cross-model requirement.\n\n## Step 5 — Completion Check\n\nAfter all targeted tasks reach Done:\n\n- If `--retro` flag was passed, or if all tasks in the feature (not just the\n  targeted subset) have reached Done for the first time in this session:\n  invoke `\u002Fsdd-quality-loop:workflow-retrospective specs\u002F\u003Cfeature>`.\n- Print a summary: number of tasks implemented, number Done, any remaining\n  non-Done tasks (with their current Status).\n\n## State Machine Summary\n\n```\nIDLE\n→ TARGET_SELECTED    (path argument or zero-arg auto-select)\n→ TRACK_DETECTED     (print [sdd-ship] Track: ... message)\n→ IMPLEMENTING       (implement-tasks for full \u002F implement-task loop for lite)\n→ QUALITY_GATE_LOOP  (quality-gate or lite-gate per task)\n   ├── task Done → continue loop\n   ├── task Blocked → STOP (surface review tickets)\n   └── ≥3 disk gate reports for task and not Done → STOP (Escalate-Human)\n       └── explicit applicable human decision → documented scoped continuation\n→ COMPLETION_CHECK   (all targeted tasks Done?)\n   └── yes → [--retro or all-Done] RETROSPECTIVE → DONE\n```\n\n## Context Compaction\n\nShip sessions span many tasks. All loop state is persisted on disk (tasks.md\nStatus fields, `reports\u002Fquality-gate\u002F`, review tickets), and both the Resume\nbehaviour (Step 3) and the disk-based cycle limit (Step 4) are designed to\nsurvive re-invocation — compaction at a task boundary loses nothing.\n\nCompact (or accept auto-compaction) only at task boundaries:\n\n- after a task reaches Done (quality gate PASS)\n- after a BLOCKED stop, before starting fix-by-review-ticket\n- between Step 3 and Step 4, when all targeted tasks are at\n  Implementation Complete\n\nDo not compact while a task is In Progress or a quality-gate cycle is\nrunning: the implementing agent's working context (file paths, partial\nchanges, verification state) is not yet reflected on disk. After compacting,\nre-invoke `\u002Fsdd-ship:ship` and rely on Resume behaviour to continue from the\nfirst eligible task.\n\n## Security Boundaries\n\n- **Never** invoke `sdd-sudo` or create\u002Fmodify the `SDD_SUDO` file.\n- **Never** set `Approval: Approved` on any task. The approval guard will block\n  this; attempting it causes confusing error messages.\n- **Never** modify gate scripts (`scripts\u002F*.sh`, `scripts\u002F*.ps1`) or hook files.\n- **Never** push to remote, create pull requests, or merge branches unless the\n  user explicitly requests it in this session.\n- **Never** modify files under `plugins\u002Fsdd-quality-loop\u002Fhooks\u002F` or\n  `plugins\u002Fsdd-quality-loop\u002Fscripts\u002Fsdd-hook-guard.*`.\n- Do not exceed the scope of the targeted tasks. Do not make changes to files\n  unrelated to the approved tasks.\n\n## Handoff\n\nAfter sdd-ship completes, report:\n\n- Tasks implemented and their final status (Done \u002F Blocked)\n- Any review tickets created (with file paths)\n- Next action for the user:\n  - If all Done: \"Feature complete. Consider re-running \u002Fsdd-ship:ship --retro\n    specs\u002F\u003Cfeature>\u002Ftasks.md to capture improvements.\"\n  - If Blocked: \"Address review tickets, apply fixes with\n    \u002Fsdd-quality-loop:fix-by-review-ticket, then re-run \u002Fsdd-ship:ship.\"\n",{"name":5,"description":7,"disable-model-invocation":23},true,"main","fd25c1e5f7b94dc5e50addb995d4313da0dda65a","plugins\u002Fsdd-ship\u002Fskills\u002Fship","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge","https:\u002F\u002Fgithub.com\u002Faharada54914","2026-11-15 02:49:38",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":23},[32,35,38],{"tool":33,"command":34},"npx","npx skills add aharada54914\u002Fsdd-forge --skill ship",{"tool":36,"command":37},"bunx","bunx skills add aharada54914\u002Fsdd-forge --skill ship",{"tool":39,"command":40},"pnpm","pnpm dlx skills add aharada54914\u002Fsdd-forge --skill ship","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fship\u002FSKILL.md",[],[44,52,59,66,73,78],{"skillNo":45,"name":46,"nameZh":6,"description":47,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":48,"categories":49,"keywords":50,"publishedAt":51},"01M7DY6PWMGRM0GFNBXAE2RP3K","status","Show a read-only Jira-style six-view SDD dashboard in the initiating parent's Sites or Claude Artifacts, with source freshness and failure visibility.","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fstatus\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M75FN16WRENJ2A9CXZCXVKBQ","verify-review-reservation","Recheck the bounded identity-ledger reservation invariant after lock, hash-check or publication changes.","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fproject-skills\u002Fformal-review-reservation\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":18},"01M75FMW90YZDCB7QNJ8K5KGKG","aws-cdk","Authors, deploys, and troubleshoots AWS infrastructure using CDK with TypeScript or Python. Covers best practices, stack architecture, and construct patterns. Applies when writing CDK constructs, bootstrapping environments, running cdk deploy\u002Fsynth\u002Fdiff, fixing CDK or CloudFormation errors, planning stack structure, importing existing resources, resolving drift, or refactoring stacks without resource replacement.","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcloud-orchestration\u002Faws-bundle\u002Fupstream\u002Fplugins\u002Faws-core\u002Fskills\u002Faws-cdk\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M75FMSTFNT7EQSZY81T4WRHN","quality-gate","Independently verify an Implementation Complete task with a Default-FAIL verification contract, deterministic checks, and an isolated critical reviewer, then decide whether it is Done.","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fspecs\u002Fsdd-domain-multitarget\u002Fverification\u002FT-003\u002Fattempt10-20261016\u002Fbaseline\u002Fplugins\u002Fsdd-quality-loop\u002Fskills\u002Fquality-gate\u002FSKILL.md",[],[],{"skillNo":74,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M75FMPM8FDE4NQ6FFY7KB83Y","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fspecs\u002Fquality-loop-fixes\u002Fhuman-copy\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fship\u002FSKILL.md",[],[],{"skillNo":79,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":18},"01M75FMK3G0EPXW77XK8ED3F2E","https:\u002F\u002Fgithub.com\u002Faharada54914\u002Fsdd-forge\u002Fblob\u002Fmain\u002Fspecs\u002Fmcp-readonly-preflight\u002Fhuman-copy\u002Fplugins\u002Fsdd-ship\u002Fskills\u002Fship\u002FSKILL.md",[],[]]