[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79DSFAEESERP1ZEXCGYTAS4":3,"skill-same:01M79DSFAEESERP1ZEXCGYTAS4":93},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":47,"commitSha":48,"skillDirectory":49,"forksCount":14,"repositoryHtmlUrl":50,"repositoryHomepageUrl":6,"ownerHtmlUrl":51,"ownerBlogUrl":6,"ownerDescription":52,"updatedAt":53,"install":54,"resources":66},"01M79DSFAEESERP1ZEXCGYTAS4","harness-flow",null,"End-to-end operator-driven flow from a JIRA\u002FConfluence requirement to a reviewed, committed (never pushed) change. Ingests an issue key or Confluence URL, verifies the requirement, asks the BA back via a ticket comment when unclear, plans, splits into worktrees, works, reviews, gets operator confirmation, then creates commits (the operator pushes manually). Trigger: run the flow, ingest a requirement, issue key, PROJ-123, confluence page, requirement to commit. Do NOT load for: standalone planning, review, or release.","foden303\u002Fharness","foden303","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F77870743?v=4","01M79DK06BJVG7HTKT05FM221X","Foden Ngo","01M79DK071J1HMYV20FJSMG71X",0,"MIT","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: harness-flow\ndescription: \"End-to-end operator-driven flow from a JIRA\u002FConfluence requirement to a reviewed, committed (never pushed) change. Ingests an issue key or Confluence URL, verifies the requirement, asks the BA back via a ticket comment when unclear, plans, splits into worktrees, works, reviews, gets operator confirmation, then creates commits (the operator pushes manually). Trigger: run the flow, ingest a requirement, issue key, PROJ-123, confluence page, requirement to commit. Do NOT load for: standalone planning, review, or release.\"\ndescription-en: \"End-to-end operator-driven flow from a JIRA\u002FConfluence requirement to a reviewed, committed (never pushed) change. Ingests an issue key or Confluence URL, verifies the requirement, asks the BA back via a ticket comment when unclear, plans, splits into worktrees, works, reviews, gets operator confirmation, then creates commits (the operator pushes manually). Trigger: run the flow, ingest a requirement, issue key, PROJ-123, confluence page, requirement to commit. Do NOT load for: standalone planning, review, or release.\"\nkind: workflow\npurpose: \"Orchestrate requirement ingest -> verify -> BA loop -> plan -> worktree -> work -> review -> confirm -> commit (no push)\"\ntrigger: \"run the flow, ingest a requirement, issue key, confluence url\"\nshape: delegate\nrole: orchestrator\nbase: harness-work\npair: harness-plan\nowner: harness-core\nsince: \"2026-07-16\"\nallowed-tools: [\"Read\", \"Write\", \"Edit\", \"Bash\", \"Grep\", \"Glob\", \"Task\", \"AskUserQuestion\", \"ScheduleWakeup\", \"mcp__claude_ai_Atlassian_Rovo__*\"]\nargument-hint: \"[\u003CISSUE-KEY>|\u003Curl> ...] [--resume \u003Csession-id>] [--accept-html] [--close-jira] [--dry-run]\"\nuser-invocable: true\neffort: high\n---\n\n# harness-flow\n\nA thin orchestrator that walks one real ticket end to end:\n\n> **JIRA\u002FConfluence requirement → verify → (ask the BA back if unclear) → plan →\n> worktree → work → review → operator confirms → commit → (operator pushes manually).**\n\nEvery heavy stage is delegated to an existing skill (`harness-plan`,\n`harness-work`, `harness-review`) — this skill only ingests the external\nrequirement, runs the BA-clarification loop, gates on operator confirmation, and\ncreates the final commits. **It never pushes.**\n\n## Core contract (read first)\n\n- **Operator supplies the target.** Trigger with a JIRA issue key (`PROJ-123`) or\n  a Confluence page URL. There is no JQL\u002Fboard polling.\n- **One link = one ticket; multiple links = one merged feature.** Passing several\n  refs (`PROJ-123 PROJ-124 PROJ-125`) means they are the **same feature** —\n  harness-flow ingests all of them, **merges into a single requirement**, and\n  verifies\u002Fplans\u002Fconfirms\u002Fcommits them as one unit (one `flow-session`). It does\n  not run them as independent tickets. The plan may still split the merged\n  feature internally into phases\u002Fsub-plans.\n- **The harness commits, the operator pushes.** harness-flow never runs\n  `git push`, `gh pr create`, or any remote transmission. See\n  [`${CLAUDE_SKILL_DIR}\u002Freferences\u002Fcommit-close.md`](${CLAUDE_SKILL_DIR}\u002Freferences\u002Fcommit-close.md).\n- **Every external write needs your approval first.** Posting a BA comment,\n  transitioning a JIRA issue, or posting a done comment are external\n  transmissions — harness-flow drafts them, shows them to you, and only sends on\n  an explicit approval. Nothing is auto-posted to JIRA\u002FConfluence.\n- **Re-entrant and idempotent.** State lives in `flow-session.v1`; every\n  invocation resumes at the recorded `status` and never repeats a completed step.\n- **MCP dependency.** JIRA\u002FConfluence access uses the Atlassian Rovo MCP tools\n  (`mcp__claude_ai_Atlassian_Rovo__*`). If they are absent (headless\u002Fcron), the\n  tri-state health probe reports `not-configured` and the skill pauses at the\n  first step that needs them rather than guessing.\n\n## Quick Reference\n\n| Input | Behavior |\n|------|------|\n| `\u002Fharness-flow PROJ-123` | Ingest JIRA issue PROJ-123 and run the full flow |\n| `\u002Fharness-flow PROJ-123 PROJ-124 PROJ-125` | Treat all three as ONE feature — merge, verify, plan, commit together |\n| `\u002Fharness-flow https:\u002F\u002F\u003Csite>\u002Fwiki\u002F...\u002FpageId` | Ingest a Confluence page and run the full flow |\n| `\u002Fharness-flow --resume \u003Csession-id>` | Resume a paused run (BA reply arrived, or manual continue) |\n| `\u002Fharness-flow PROJ-123 --accept-html` | Use the harness-accept ship\u002Fwait\u002Freject HTML at the confirm gate |\n| `\u002Fharness-flow PROJ-123 --close-jira` | After commit, transition the issue + post a done comment |\n| `\u002Fharness-flow PROJ-123 --dry-run` | Walk the state machine with no MCP writes \u002F no git commits |\n\n## Options\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `\u003CISSUE-KEY>` \u002F `\u003Cconfluence-url>` | The requirement source | required (unless `--resume`) |\n| `--resume \u003Csession-id>` | Resume a paused session | - |\n| `--accept-html` | Render harness-accept HTML at the confirm gate | off (uses `decision_needed.v1`) |\n| `--close-jira` | Transition the issue + post a done comment after commit | off |\n| `--dry-run` | No external writes, no commits | off |\n\n## Resolving the plugin bundle root\n\nharness-flow calls helper scripts under the plugin bundle root, not the host\nproject's cwd. At the start of every invocation resolve `HARNESS_PLUGIN_ROOT`:\n\n1. If `CLAUDE_PLUGIN_ROOT` exists and contains `scripts\u002F`, use it.\n2. Otherwise derive it from `CLAUDE_SKILL_DIR`\n   (`skills\u002Fharness-flow` → `${CLAUDE_SKILL_DIR}\u002F..\u002F..`;\n   `.agents\u002Fskills\u002Fharness-flow` mirror → `${CLAUDE_SKILL_DIR}\u002F..\u002F..\u002F..`).\n3. If neither resolves, stop and re-run after setting `CLAUDE_PLUGIN_ROOT`.\n\nKeep `Plans.md` and `.claude\u002Fstate\u002F...` on the host project side. Call only\nhelper scripts from `${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002F...`.\n\n## State machine\n\nState file: `.claude\u002Fstate\u002Fflow\u002F\u003Csession-id>\u002Fsession.json` (`flow-session.v1`),\nmanaged only through `${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002Fflow-session.sh`.\n`\u003Csession-id>` is derived from the source ref (e.g. `PROJ-123` → `proj-123`;\nConfluence page id → `conf-\u003CpageId>`), so re-running the same target resumes the\nsame session.\n\nOn every invocation:\n\n1. Resolve `HARNESS_PLUGIN_ROOT` (above).\n2. If `--resume \u003Cid>` or a session file already exists for the target, read it\n   and jump to the handler for its `status`. Otherwise start at Step 1.\n3. Probe MCP health once and record it:\n   `bash \"${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002Fflow-mcp-health.sh\" --probe \u003Cresult>`\n   where `\u003Cresult>` is what a lightweight Rovo call\n   (`getAccessibleAtlassianResources`) returned: success → `healthy`,\n   call error → `unreachable`, tools not present → `not-configured`.\n   Persist with `flow-session.sh set session.json mcp_health \u003Cstate>`.\n\n```\n[Step 1] Ingest        status: ingesting -> ingested\n  Reference: references\u002Fingestion.md\n  For EACH ref passed (1 or many):\n    JIRA key   -> getJiraIssue (summary, description, acceptance-criteria field, labels, type, status, reporter)\n    Conf URL   -> getConfluencePage (+ getConfluencePageFooterComments)\n  atlassianUserInfo once -> cache bot accountId into clarification.bot_account_id\n  If >1 ref: MERGE them into ONE requirement (union of criteria, concatenated\n    descriptions with per-ticket headers) and pass --sources-file listing all\n    refs; the primary (first) ref becomes source\u002Fsource_ref + session_id.\n  Extract fields in-context, then:\n    flow-ingest-requirement.sh --source \u003Cs> --source-ref \u003Cref> --title \u003Ct> \\\n      --description-file \u003Ctmp> --acceptance-criteria-file \u003Ctmp> --labels a,b \\\n      --issue-type \u003CT> --status \u003CS> --reporter-account-id \u003Cid> \\\n      --mcp-available true --out .claude\u002Fstate\u002Fflow\u002F\u003Cid>\u002Frequirement.json\n  flow-session.sh set session.json requirement_path \u003Cpath>\n  |\n  v\n[Step 2] Verify        status: verifying\n  Reference: references\u002Fverify-rubric.md  (6 binary gates)\n  Write the verification block into requirement.json.\n  verdict == ok               -> Step 4\n  verdict == needs-clarification -> Step 3\n  |\n  v\n[Step 3] BA loop       status: awaiting-ba  \u003C-> verifying\n  Reference: references\u002Fba-loop.md\n  Draft a MARKED comment with the open_questions, then GET OPERATOR APPROVAL\n  before posting (posting = external transmission; never auto-post):\n    show the draft via AskUserQuestion -> only on \"Post it\" do you send\n    JIRA       -> addCommentToJiraIssue\n    Confluence -> createConfluenceFooterComment\n  Store clarification {nonce, posted_comment_id, posted_at, bot_account_id, rounds}.\n  AFTER the approved post, auto-poll via ScheduleWakeup (see ba-loop.md): each\n  wake re-fetches comments, matches the BA reply, folds it into requirement.json,\n  re-runs Step 2. (Auto-poll only READS; no approval needed to read.)\n  Round cap (3) -> manual --resume -> escalate via decision_needed.v1.\n  MCP unreachable here -> emit decision_needed.v1 and pause.\n  |\n  v\n[Step 4] Plan          status: planning\n  Reference: references\u002Fplanning-split.md\n  Delegate to harness-plan create with the requirement + JIRA key as provenance.\n  Split long requirements into Phases + numbered subtasks (preferred), or named\n  plans for genuinely independent deliverables (record in session.plans[]).\n  Persist the ## Pre-approval block to .claude\u002Fstate\u002Fplan-preapprovals.json.\n  Gate: go run .\u002Fcmd\u002Fharness plans check-deps\n  |\n  v\n[Step 5-6] Worktree + Work   status: working\n  Delegate to harness-work all (auto: 1=Solo \u002F 2-3=Parallel \u002F 4+=Breezing).\n  Worktrees (.harness-worktrees\u002F), sprint contracts, TDD RED logs, and the\n  worker self_review[] gate all run inside harness-work.\n  Inject the JIRA key into the worker\u002Flead briefing so commit messages carry\n  [PROJ-123]. Record produced commit hashes into session.commit_hashes[].\n  |\n  v\n[Step 7] Review        status: reviewing\n  Reuse harness-work's built-in review loop (review-result.v1). For high-risk\n  labels detected at ingestion, add harness-review --dual \u002F --security.\n  |\n  v\n[Step 8] Confirm       status: awaiting-confirm\n  Reference: references\u002Fconfirm-gate.md\n  Default: decision_needed.v1 CLI gate (requirement + JIRA key, per-criterion\n  verified\u002Funverified, verdict, changed files, commit-message preview).\n  --accept-html: render harness-accept ship\u002Fwait\u002Freject.\n  OK        -> Step 9\n  not-OK    -> Step 10 (capture the reason)\n  |\n  v\n[Step 9] Commit        status: committing -> done\n  Reference: references\u002Fcommit-close.md\n  Per-task commits, each tagged [PROJ-123]. NEVER push.\n  --close-jira: getTransitionsForJiraIssue -> transitionJiraIssue + a done comment\n  noting the change is committed locally and awaits the operator's manual push.\n  |\n  v\n[Step 10] Fix \u002F rework   status: working (re-entry)\n  Route not-OK \u002F REQUEST_CHANGES back to harness-work on the failing task(s),\n  reusing references\u002Ffailure-reticketing.md (per-task counter, buildFixTaskID,\n  3-strike escalation). The operator's not-OK reason becomes the fix-task Content.\n  After APPROVE, re-enter Step 8. Cap rework_rounds; escalate on exhaustion.\n```\n\n## Stop \u002F pause conditions\n\n| Condition | status | Response |\n|-----------|--------|----------|\n| BA reply not yet available | `awaiting-ba` | Auto-poll (ScheduleWakeup) or wait for `--resume` |\n| Clarification round cap reached | `escalated` | `decision_needed.v1` (request-unachievable) |\n| MCP unreachable at a step that needs it | unchanged | `decision_needed.v1`, pause |\n| Operator answers not-OK at confirm | `working` | Fix\u002Frework (Step 10) |\n| Rework round cap reached | `escalated` | Escalate to the operator |\n| All committed | `done` | Report commit hashes + remind operator to push manually |\n\n## Silence policy\n\nReport on: ingest done, verification verdict, BA comment posted, BA reply folded\nin, plan created, work\u002Freview verdict per task, confirm-gate result, commit\nhashes, escalation. Stay silent on: pacing waits between BA-loop wakes with no\nnew reply, fine-grained stdout. Always end a `done` run by naming the commit\nhashes and stating the branch is **un-pushed** (operator pushes manually).\n\n## Related skills\n\n- `harness-plan` — planning (Step 4), invoked via `create`\n- `harness-work` — worktree + implementation + review loop (Steps 5-7)\n- `harness-review` — optional extra review lens for high-risk changes (Step 7)\n- `harness-accept` — optional ship\u002Fwait\u002Freject HTML at the confirm gate (Step 8)\n- `harness-loop` — the ScheduleWakeup re-entry pattern reused by the BA loop\n","# harness-flow\n\nA thin orchestrator that walks one real ticket end to end:\n\n> **JIRA\u002FConfluence requirement → verify → (ask the BA back if unclear) → plan →\n> worktree → work → review → operator confirms → commit → (operator pushes manually).**\n\nEvery heavy stage is delegated to an existing skill (`harness-plan`,\n`harness-work`, `harness-review`) — this skill only ingests the external\nrequirement, runs the BA-clarification loop, gates on operator confirmation, and\ncreates the final commits. **It never pushes.**\n\n## Core contract (read first)\n\n- **Operator supplies the target.** Trigger with a JIRA issue key (`PROJ-123`) or\n  a Confluence page URL. There is no JQL\u002Fboard polling.\n- **One link = one ticket; multiple links = one merged feature.** Passing several\n  refs (`PROJ-123 PROJ-124 PROJ-125`) means they are the **same feature** —\n  harness-flow ingests all of them, **merges into a single requirement**, and\n  verifies\u002Fplans\u002Fconfirms\u002Fcommits them as one unit (one `flow-session`). It does\n  not run them as independent tickets. The plan may still split the merged\n  feature internally into phases\u002Fsub-plans.\n- **The harness commits, the operator pushes.** harness-flow never runs\n  `git push`, `gh pr create`, or any remote transmission. See\n  [`${CLAUDE_SKILL_DIR}\u002Freferences\u002Fcommit-close.md`](${CLAUDE_SKILL_DIR}\u002Freferences\u002Fcommit-close.md).\n- **Every external write needs your approval first.** Posting a BA comment,\n  transitioning a JIRA issue, or posting a done comment are external\n  transmissions — harness-flow drafts them, shows them to you, and only sends on\n  an explicit approval. Nothing is auto-posted to JIRA\u002FConfluence.\n- **Re-entrant and idempotent.** State lives in `flow-session.v1`; every\n  invocation resumes at the recorded `status` and never repeats a completed step.\n- **MCP dependency.** JIRA\u002FConfluence access uses the Atlassian Rovo MCP tools\n  (`mcp__claude_ai_Atlassian_Rovo__*`). If they are absent (headless\u002Fcron), the\n  tri-state health probe reports `not-configured` and the skill pauses at the\n  first step that needs them rather than guessing.\n\n## Quick Reference\n\n| Input | Behavior |\n|------|------|\n| `\u002Fharness-flow PROJ-123` | Ingest JIRA issue PROJ-123 and run the full flow |\n| `\u002Fharness-flow PROJ-123 PROJ-124 PROJ-125` | Treat all three as ONE feature — merge, verify, plan, commit together |\n| `\u002Fharness-flow https:\u002F\u002F\u003Csite>\u002Fwiki\u002F...\u002FpageId` | Ingest a Confluence page and run the full flow |\n| `\u002Fharness-flow --resume \u003Csession-id>` | Resume a paused run (BA reply arrived, or manual continue) |\n| `\u002Fharness-flow PROJ-123 --accept-html` | Use the harness-accept ship\u002Fwait\u002Freject HTML at the confirm gate |\n| `\u002Fharness-flow PROJ-123 --close-jira` | After commit, transition the issue + post a done comment |\n| `\u002Fharness-flow PROJ-123 --dry-run` | Walk the state machine with no MCP writes \u002F no git commits |\n\n## Options\n\n| Option | Description | Default |\n|--------|-------------|---------|\n| `\u003CISSUE-KEY>` \u002F `\u003Cconfluence-url>` | The requirement source | required (unless `--resume`) |\n| `--resume \u003Csession-id>` | Resume a paused session | - |\n| `--accept-html` | Render harness-accept HTML at the confirm gate | off (uses `decision_needed.v1`) |\n| `--close-jira` | Transition the issue + post a done comment after commit | off |\n| `--dry-run` | No external writes, no commits | off |\n\n## Resolving the plugin bundle root\n\nharness-flow calls helper scripts under the plugin bundle root, not the host\nproject's cwd. At the start of every invocation resolve `HARNESS_PLUGIN_ROOT`:\n\n1. If `CLAUDE_PLUGIN_ROOT` exists and contains `scripts\u002F`, use it.\n2. Otherwise derive it from `CLAUDE_SKILL_DIR`\n   (`skills\u002Fharness-flow` → `${CLAUDE_SKILL_DIR}\u002F..\u002F..`;\n   `.agents\u002Fskills\u002Fharness-flow` mirror → `${CLAUDE_SKILL_DIR}\u002F..\u002F..\u002F..`).\n3. If neither resolves, stop and re-run after setting `CLAUDE_PLUGIN_ROOT`.\n\nKeep `Plans.md` and `.claude\u002Fstate\u002F...` on the host project side. Call only\nhelper scripts from `${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002F...`.\n\n## State machine\n\nState file: `.claude\u002Fstate\u002Fflow\u002F\u003Csession-id>\u002Fsession.json` (`flow-session.v1`),\nmanaged only through `${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002Fflow-session.sh`.\n`\u003Csession-id>` is derived from the source ref (e.g. `PROJ-123` → `proj-123`;\nConfluence page id → `conf-\u003CpageId>`), so re-running the same target resumes the\nsame session.\n\nOn every invocation:\n\n1. Resolve `HARNESS_PLUGIN_ROOT` (above).\n2. If `--resume \u003Cid>` or a session file already exists for the target, read it\n   and jump to the handler for its `status`. Otherwise start at Step 1.\n3. Probe MCP health once and record it:\n   `bash \"${HARNESS_PLUGIN_ROOT}\u002Fscripts\u002Fflow-mcp-health.sh\" --probe \u003Cresult>`\n   where `\u003Cresult>` is what a lightweight Rovo call\n   (`getAccessibleAtlassianResources`) returned: success → `healthy`,\n   call error → `unreachable`, tools not present → `not-configured`.\n   Persist with `flow-session.sh set session.json mcp_health \u003Cstate>`.\n\n```\n[Step 1] Ingest        status: ingesting -> ingested\n  Reference: references\u002Fingestion.md\n  For EACH ref passed (1 or many):\n    JIRA key   -> getJiraIssue (summary, description, acceptance-criteria field, labels, type, status, reporter)\n    Conf URL   -> getConfluencePage (+ getConfluencePageFooterComments)\n  atlassianUserInfo once -> cache bot accountId into clarification.bot_account_id\n  If >1 ref: MERGE them into ONE requirement (union of criteria, concatenated\n    descriptions with per-ticket headers) and pass --sources-file listing all\n    refs; the primary (first) ref becomes source\u002Fsource_ref + session_id.\n  Extract fields in-context, then:\n    flow-ingest-requirement.sh --source \u003Cs> --source-ref \u003Cref> --title \u003Ct> \\\n      --description-file \u003Ctmp> --acceptance-criteria-file \u003Ctmp> --labels a,b \\\n      --issue-type \u003CT> --status \u003CS> --reporter-account-id \u003Cid> \\\n      --mcp-available true --out .claude\u002Fstate\u002Fflow\u002F\u003Cid>\u002Frequirement.json\n  flow-session.sh set session.json requirement_path \u003Cpath>\n  |\n  v\n[Step 2] Verify        status: verifying\n  Reference: references\u002Fverify-rubric.md  (6 binary gates)\n  Write the verification block into requirement.json.\n  verdict == ok               -> Step 4\n  verdict == needs-clarification -> Step 3\n  |\n  v\n[Step 3] BA loop       status: awaiting-ba  \u003C-> verifying\n  Reference: references\u002Fba-loop.md\n  Draft a MARKED comment with the open_questions, then GET OPERATOR APPROVAL\n  before posting (posting = external transmission; never auto-post):\n    show the draft via AskUserQuestion -> only on \"Post it\" do you send\n    JIRA       -> addCommentToJiraIssue\n    Confluence -> createConfluenceFooterComment\n  Store clarification {nonce, posted_comment_id, posted_at, bot_account_id, rounds}.\n  AFTER the approved post, auto-poll via ScheduleWakeup (see ba-loop.md): each\n  wake re-fetches comments, matches the BA reply, folds it into requirement.json,\n  re-runs Step 2. (Auto-poll only READS; no approval needed to read.)\n  Round cap (3) -> manual --resume -> escalate via decision_needed.v1.\n  MCP unreachable here -> emit decision_needed.v1 and pause.\n  |\n  v\n[Step 4] Plan          status: planning\n  Reference: references\u002Fplanning-split.md\n  Delegate to harness-plan create with the requirement + JIRA key as provenance.\n  Split long requirements into Phases + numbered subtasks (preferred), or named\n  plans for genuinely independent deliverables (record in session.plans[]).\n  Persist the ## Pre-approval block to .claude\u002Fstate\u002Fplan-preapprovals.json.\n  Gate: go run .\u002Fcmd\u002Fharness plans check-deps\n  |\n  v\n[Step 5-6] Worktree + Work   status: working\n  Delegate to harness-work all (auto: 1=Solo \u002F 2-3=Parallel \u002F 4+=Breezing).\n  Worktrees (.harness-worktrees\u002F), sprint contracts, TDD RED logs, and the\n  worker self_review[] gate all run inside harness-work.\n  Inject the JIRA key into the worker\u002Flead briefing so commit messages carry\n  [PROJ-123]. Record produced commit hashes into session.commit_hashes[].\n  |\n  v\n[Step 7] Review        status: reviewing\n  Reuse harness-work's built-in review loop (review-result.v1). For high-risk\n  labels detected at ingestion, add harness-review --dual \u002F --security.\n  |\n  v\n[Step 8] Confirm       status: awaiting-confirm\n  Reference: references\u002Fconfirm-gate.md\n  Default: decision_needed.v1 CLI gate (requirement + JIRA key, per-criterion\n  verified\u002Funverified, verdict, changed files, commit-message preview).\n  --accept-html: render harness-accept ship\u002Fwait\u002Freject.\n  OK        -> Step 9\n  not-OK    -> Step 10 (capture the reason)\n  |\n  v\n[Step 9] Commit        status: committing -> done\n  Reference: references\u002Fcommit-close.md\n  Per-task commits, each tagged [PROJ-123]. NEVER push.\n  --close-jira: getTransitionsForJiraIssue -> transitionJiraIssue + a done comment\n  noting the change is committed locally and awaits the operator's manual push.\n  |\n  v\n[Step 10] Fix \u002F rework   status: working (re-entry)\n  Route not-OK \u002F REQUEST_CHANGES back to harness-work on the failing task(s),\n  reusing references\u002Ffailure-reticketing.md (per-task counter, buildFixTaskID,\n  3-strike escalation). The operator's not-OK reason becomes the fix-task Content.\n  After APPROVE, re-enter Step 8. Cap rework_rounds; escalate on exhaustion.\n```\n\n## Stop \u002F pause conditions\n\n| Condition | status | Response |\n|-----------|--------|----------|\n| BA reply not yet available | `awaiting-ba` | Auto-poll (ScheduleWakeup) or wait for `--resume` |\n| Clarification round cap reached | `escalated` | `decision_needed.v1` (request-unachievable) |\n| MCP unreachable at a step that needs it | unchanged | `decision_needed.v1`, pause |\n| Operator answers not-OK at confirm | `working` | Fix\u002Frework (Step 10) |\n| Rework round cap reached | `escalated` | Escalate to the operator |\n| All committed | `done` | Report commit hashes + remind operator to push manually |\n\n## Silence policy\n\nReport on: ingest done, verification verdict, BA comment posted, BA reply folded\nin, plan created, work\u002Freview verdict per task, confirm-gate result, commit\nhashes, escalation. Stay silent on: pacing waits between BA-loop wakes with no\nnew reply, fine-grained stdout. Always end a `done` run by naming the commit\nhashes and stating the branch is **un-pushed** (operator pushes manually).\n\n## Related skills\n\n- `harness-plan` — planning (Step 4), invoked via `create`\n- `harness-work` — worktree + implementation + review loop (Steps 5-7)\n- `harness-review` — optional extra review lens for high-risk changes (Step 7)\n- `harness-accept` — optional ship\u002Fwait\u002Freject HTML at the confirm gate (Step 8)\n- `harness-loop` — the ScheduleWakeup re-entry pattern reused by the BA loop\n",{"base":24,"kind":25,"name":5,"pair":26,"role":27,"owner":28,"shape":29,"since":30,"effort":31,"purpose":32,"trigger":33,"description":7,"allowed-tools":34,"argument-hint":45,"description-en":7,"user-invocable":46},"harness-work","workflow","harness-plan","orchestrator","harness-core","delegate","2026-07-16","high","Orchestrate requirement ingest -> verify -> BA loop -> plan -> worktree -> work -> review -> confirm -> commit (no push)","run the flow, ingest a requirement, issue key, confluence url",[35,36,37,38,39,40,41,42,43,44],"Read","Write","Edit","Bash","Grep","Glob","Task","AskUserQuestion","ScheduleWakeup","mcp__claude_ai_Atlassian_Rovo__*","[\u003CISSUE-KEY>|\u003Curl> ...] [--resume \u003Csession-id>] [--accept-html] [--close-jira] [--dry-run]",true,"main","3517f6fc9c5159f32d585682386ed45e35c30ddf","skills\u002Fharness-flow","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness","https:\u002F\u002Fgithub.com\u002Ffoden303","Software Engineer","2026-11-06 10:59:21",{"source":8,"skillName":5,"commands":55,"prompt":65,"zipAvailable":46},[56,59,62],{"tool":57,"command":58},"npx","npx skills add foden303\u002Fharness --skill harness-flow",{"tool":60,"command":61},"bunx","bunx skills add foden303\u002Fharness --skill harness-flow",{"tool":63,"command":64},"pnpm","pnpm dlx skills add foden303\u002Fharness --skill harness-flow","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002FSKILL.md",[67,73,77,81,85,89],{"path":68,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fba-loop.md","REFERENCE","参考文档",5380,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fba-loop.md",{"path":74,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Fcommit-close.md",4588,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fcommit-close.md",{"path":78,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":79,"sourceHtmlUrl":80},"references\u002Fconfirm-gate.md",2521,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fconfirm-gate.md",{"path":82,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":83,"sourceHtmlUrl":84},"references\u002Fingestion.md",5796,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fingestion.md",{"path":86,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":87,"sourceHtmlUrl":88},"references\u002Fplanning-split.md",3486,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fplanning-split.md",{"path":90,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":91,"sourceHtmlUrl":92},"references\u002Fverify-rubric.md",3081,"https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-flow\u002Freferences\u002Fverify-rubric.md",[94,100,106,113,120,126],{"skillNo":95,"name":96,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":19},"01M79DWR27RWW11HNKY0MRECPP","escape-skill","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fzz-review-escape\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":19},"01M79DWPSR8ZK3V4DQK5EWVJ3K","zz-review-empty","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fskills\u002Fzz-review-empty\u002FSKILL.md",[],[],{"skillNo":107,"name":108,"nameZh":6,"description":109,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":19},"01M79DWFV6PYV0QEZ3SC32VMCM","memory","Manage SSOT, memory, and cross-tool memory search. Guardian of decisions.md and patterns.md. Use when user mentions memory, SSOT, decisions.md, patterns.md, merging, migration, SSOT promotion, sync memory, save learnings, memory search, harness-mem, past decisions, or record this. Do NOT load for: implementation work, reviews, ad-hoc notes, or in-session logging.","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fmemory\u002FSKILL.md",[],[],{"skillNo":114,"name":115,"nameZh":6,"description":116,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":119,"publishedAt":19},"01M79DWB6MB7EQK2JT4A84NHJD","maintenance","File cleanup and archiving. Tidies up bloated Plans.md, session-log.md, old logs, and state files. Trigger: \u002Fmaintenance, cleanup, archive, organize, split session-log. Do NOT load for: implementation, review, release, new feature development.","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fmaintenance\u002FSKILL.md",[],[],{"skillNo":121,"name":24,"nameZh":6,"description":122,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":123,"categories":124,"keywords":125,"publishedAt":19},"01M79DW33AEPBCXFAMX8YH56PG","HAR: Execute Plans.md tasks from single task to full parallel team run. Trigger: implement, execute, do everything, breezing, team run, parallel. Do NOT load for: planning, review, release, setup.","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-work\u002FSKILL.md",[],[],{"skillNo":127,"name":128,"nameZh":6,"description":129,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":130,"categories":131,"keywords":132,"publishedAt":19},"01M79DVWV7PZC0E2GYC103QBYZ","harness-sync","HAR: Sync Plans.md with implementation. Drift detect, marker update, retrospective. Trigger: sync-status, where am I, check progress. --snapshot for snapshots. Do NOT load for: planning, implementation, review, release.","https:\u002F\u002Fgithub.com\u002Ffoden303\u002Fharness\u002Fblob\u002Fmain\u002Fskills\u002Fharness-sync\u002FSKILL.md",[],[]]