[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6QEB1MG3NCEZ9DP310AAANC":3,"skill-same:01M6QEB1MG3NCEZ9DP310AAANC":67},{"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":23,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":44,"commitSha":45,"skillDirectory":46,"forksCount":47,"repositoryHtmlUrl":48,"repositoryHomepageUrl":49,"ownerHtmlUrl":50,"ownerBlogUrl":6,"ownerDescription":51,"updatedAt":52,"install":53,"resources":66},"01M6QEB1MG3NCEZ9DP310AAANC","gitguardex",null,"Use for GitGuardex branch\u002Fworktree safety, agent or subagent collision avoidance, file ownership, and gated PR finishing. Not for code-quality review.","opencue\u002Fgitguardex","opencue","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F271874627?v=4","01M6QDWWMZFB8A08PDTA8D0VJ3","Opencue","01M6QDWWSJGBEGVPMCEJFWVACT",28,"MIT","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Ftemplates\u002Fcodex\u002Fskills\u002Fgitguardex\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[5,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"review","pr","code","worktree","file","Safety","code-quality","subagent","quality","branch","use","gated","quality review","ownership","for","2026-11-10 10:12:30",false,"---\nname: gitguardex\ndescription: \"Use for GitGuardex branch\u002Fworktree safety, agent or subagent collision avoidance, file ownership, and gated PR finishing. Not for code-quality review.\"\n---\n\nUse when a GitGuardex-managed repository needs safe edits or multi-agent coordination.\n\n`gx status` -> `gx doctor` -> `gx status --strict`\n\nBootstrap: `gx setup`\nOps: `gx branch start \"\u003Ctask>\" \"\u003Cagent>\"`, `gx locks claim --branch \"\u003Cagent-branch>\" \u003Cfile...>`, `gx branch finish --branch \"\u003Cagent-branch>\" --base \u003Cbase> --via-pr --wait-for-merge --cleanup`, `gx finish --all`, `gx cleanup`\n\n## Choose the lane before editing\n\nWorktree policy is strict `always` unless the repo opts into `adaptive` with\n`GUARDEX_WORKTREE_MODE=adaptive` in `.env` or\n`git config --local multiagent.worktreeMode adaptive`.\n\n1. Run `gx status`. In adaptive mode, also inspect `list_agents` (or\n   `gx mcp list-agents --no-prs`), the current dirty paths, and `who_owns` for\n   every intended file.\n2. Stay on the current checkout only for bounded, single-agent work with no\n   competing writer, claimed target, or dirty-path overlap.\n3. Pivot to a new isolated lane when the task is substantial\u002Flong-lived,\n   another writer is active in the repo, a target is owned or dirty elsewhere,\n   or the scope expands. Use\n   `gx branch start --new --no-transfer \"\u003Cbounded task>\" \"\u003Cagent>\"`.\n4. Re-run the radar before expanding scope. Never overwrite another lane's\n   changes. If direct work already has edits when a pivot becomes necessary,\n   transfer only those known edits deliberately; do not absorb unrelated files.\n\nDirect adaptive work supports file edits, `git add`\u002Fordinary commit\u002Fpush, and\nbounded test\u002Flint\u002Fbuild commands. Custom executors, scripts, or history rewrites\npivot to an isolated lane. Isolated work uses claims and the gated\n`gx branch finish` flow below.\n\n## Keep the default workflow lean\n\n- Default to direct T1 work in code and tests. Do not create OpenSpec artifacts from task wording alone.\n- OpenSpec is explicit opt-in: use `--tier T2` for a change workspace or `--tier T3` for change plus plan only when the user requests OpenSpec or the caller explicitly selects T2\u002FT3.\n- Treat the commit, targeted test output, and compact final handoff as the routine task record.\n\n## Agent and subagent collaboration\n\nBefore parallel work, inspect the field with the MCP `list_agents` tool or the CLI fallback:\n\n```sh\ngx mcp list-agents --no-prs\n```\n\nFor every file a parallel writer may touch, check ownership, create an isolated lane, then claim it:\n\n```sh\ngx mcp who-owns path\u002Fto\u002Ffile.ts\ngx branch start --new --no-transfer \"\u003Cbounded task>\" \"\u003Cagent>\"\ngx locks claim --branch \"\u003Cagent-branch>\" path\u002Fto\u002Ffile.ts\n```\n\n- **One owner per file.** Split writer subagents by non-overlapping files or modules. Read-only scouts may share a surface; writers may not.\n- Give each subagent one bounded objective and a verification command. Do not ask one subagent to explore, redesign, implement, and review the same slice.\n- Re-check `list_agents` before expanding scope. If another lane owns or is already editing a file, coordinate a handoff or choose a different slice.\n- Return a compact handoff: conclusion, `file:line`, changed files, verification result, and blocker or next action. Do not relay full logs.\n- Finish each writing lane through its own PR once. Never run duplicate finish commands for the same branch.\n\n## Optional cross-machine Git lock and radar state\n\nLocal worktree locks remain the default. When agents on multiple machines use the same writable Git remote, opt into shared state per repository:\n\n```sh\ngx locks shared-enable --remote origin\ngx locks shared-status\n```\n\nAfter enablement, ordinary `claim`, `allow-delete`, `validate`, `release`, and `status` commands also use atomic remote refs under `refs\u002Fgitguardex\u002Flocks\u002F*`. `gx mcp list-agents` combines local worktrees with remote `agent\u002F*` branches and shared lock owners; `gx mcp who-owns` resolves the same shared owner data.\n\n- Shared mode is explicit and fail-closed: if the remote is unreachable, rejects custom refs, or returns invalid metadata, do not keep a local-only claim or approve a commit.\n- The remote must allow authenticated custom-ref creation, compare-and-swap updates, and deletion. GitGuardex stores validated JSON metadata in commits; it never executes remote metadata, and publishes only a hashed machine identifier.\n- Push the agent branch when full remote-lane visibility matters. A lock still makes an unpushed branch visible as a lock-owning remote lane.\n- Release locks before disabling shared state or abandoning a lane. Cross-machine locks are not automatically reaped because one host cannot safely prove that another host's process is dead.\n- `gx locks shared-disable` stops shared reads and writes for this repository; it does not delete remote refs owned by any machine.\n\n## Finish checklist in Codex\n\nBefore starting a long `gx finish`, `gx ship`, or `gx branch finish`, show this compact Markdown checklist in commentary and update it from the CLI's `[gx:finish]` transition lines:\n\n- [ ] Prepare branch\n- [ ] Local preflight\n- [ ] Push and open PR\n- [ ] AI review\n- [ ] Review autofix\n- [ ] CI checks\n- [ ] Merge\n- [ ] Cleanup\n\nRender each state-change update as a compact card, not as a raw log dump:\n\n### ? GX Finish · `3\u002F8`\n`agent\u002F...` → `main` · ? **AI review** · `00:44`\n\n- ✅ Prepare branch\n- ✅ Local preflight\n- ✅ Push and open PR · `#123`\n- ? AI review\n- ⬜ Review autofix\n- ⬜ CI checks\n- ⬜ Merge\n- ⬜ Cleanup\n\nStart the command with a short initial yield, then poll the background process every 15-30 seconds. Between polls, post an updated checklist when the active stage changes; during a quiet long stage, name the active stage and elapsed time at most every 30 seconds. Never leave the user with only Codex's generic **Working** or **Waiting for background terminal** label. Do not launch a duplicate finish while the first process is still alive, and do not mark a checkbox complete until a matching `✓` line or equivalent terminal evidence arrives.\n\nTreat `? Cleanup` as finished-but-best-effort, not as unconditional success: surface any cleanup warnings beside it instead of converting it to `✅`.\n\nStructured phase events are persisted privately under `.omx\u002Fstate\u002Ffinish-runs\u002F*.jsonl` (`0700` directory, `0600` files). Prefer that JSONL stream over parsing narrative logs when exact state is needed; the visual output prints the active event-file path.\n\nFor agent transcripts, opt into `gx branch finish --agent-quiet` (or `gx finish --agent-quiet`)\nalongside the same required finish flags and gates. It prints one run ID\u002Fevent-file pointer,\nretains warnings and blockers, and ends with a bounded summary; it does not skip verification.\nRead only new observations with `gx finish events --run \u003Crun-id> --wait-ms 30000`, then\npass the returned `--cursor \u003Ccursor>` on the next read. Use the same repository target as\nthe finish invocation. Let this bounded CLI wait do the polling instead of repeatedly\nasking the model to reread the whole log. Events are observations, never merge authority.\n\nOpt-in bounded observations (defaults remain unchanged):\n- MCP `my_context`: pass `max_bytes` to bound UTF-8 JSON tool text. Ownership,\n  conflicts and safety fields are mandatory; `budget-too-small` requires a larger\n  budget or smaller file batch. Check `complete`\u002F`omitted`; fetch full peers only\n  when needed. This budgets bytes, not model tokens.\n- `gx watch --pr-cache-ms 5000` shares PR probes with a short in-process TTL keyed\n  by repository, remote, branch and HEAD. Observation age is shown; dirty state\n  stays fresh. The cache is never used by merge gates.\n- `gx finish events --run \u003Cid> --view state`: merge `stages` patches by stage,\n  retain every `alerts` record, and resume with the opaque cursor. Paginated\n  snapshots are incomplete until `snapshotComplete: true`. Raw view remains the\n  default; reset explicitly on invalid\u002Freplaced\u002Ftruncated cursors.\n- `gx agents set-status --activity working --dedupe-surface`: skip identical\n  successful pane-label writes, but still renew session activity on every call.\n  Changed targets and failed writes retry. This never infers agent completion.\n\nFor setup guidance, use `gx prompt --list-parts` then `gx prompt --part \u003Cname>` for the\nneeded slice; use `--exec --part \u003Cname>` only for command-capable slices. Do not repeatedly\nload the full setup prompt or remove the mandatory safety contract to save tokens.\n\nWhen inspecting or verifying, prefer `rtk` compact wrappers if available (`rtk git status`, `rtk grep`, `rtk test \u003Ccmd>`, and noisy gx reads like `rtk gx status` \u002F `rtk gx doctor`). Do not wrap commands whose stdout is parsed by scripts (`--json`, `--porcelain`, exact stdout contracts) or shell-ready output (`gx prompt --exec`).\n\nTo shrink gx's own large narrative output (e.g. `gx prompt`, `gx prompt --snippet`) before it lands in your context, set `GUARDEX_COMPRESS_CMD=\"\u003Cstdin->stdout filter>\"`; gx routes that output through the filter (terse\u002Fnon-TTY mode, fail-open, JSON skipped). Unset = byte-for-byte unchanged. Confirm it is wired with `gx status` or `gx doctor` — both print a `Token compression` line and flag a configured-but-missing binary (doctor's warning is advisory and never changes its safe\u002Funsafe exit code).\n","Use when a GitGuardex-managed repository needs safe edits or multi-agent coordination.\n\n`gx status` -> `gx doctor` -> `gx status --strict`\n\nBootstrap: `gx setup`\nOps: `gx branch start \"\u003Ctask>\" \"\u003Cagent>\"`, `gx locks claim --branch \"\u003Cagent-branch>\" \u003Cfile...>`, `gx branch finish --branch \"\u003Cagent-branch>\" --base \u003Cbase> --via-pr --wait-for-merge --cleanup`, `gx finish --all`, `gx cleanup`\n\n## Choose the lane before editing\n\nWorktree policy is strict `always` unless the repo opts into `adaptive` with\n`GUARDEX_WORKTREE_MODE=adaptive` in `.env` or\n`git config --local multiagent.worktreeMode adaptive`.\n\n1. Run `gx status`. In adaptive mode, also inspect `list_agents` (or\n   `gx mcp list-agents --no-prs`), the current dirty paths, and `who_owns` for\n   every intended file.\n2. Stay on the current checkout only for bounded, single-agent work with no\n   competing writer, claimed target, or dirty-path overlap.\n3. Pivot to a new isolated lane when the task is substantial\u002Flong-lived,\n   another writer is active in the repo, a target is owned or dirty elsewhere,\n   or the scope expands. Use\n   `gx branch start --new --no-transfer \"\u003Cbounded task>\" \"\u003Cagent>\"`.\n4. Re-run the radar before expanding scope. Never overwrite another lane's\n   changes. If direct work already has edits when a pivot becomes necessary,\n   transfer only those known edits deliberately; do not absorb unrelated files.\n\nDirect adaptive work supports file edits, `git add`\u002Fordinary commit\u002Fpush, and\nbounded test\u002Flint\u002Fbuild commands. Custom executors, scripts, or history rewrites\npivot to an isolated lane. Isolated work uses claims and the gated\n`gx branch finish` flow below.\n\n## Keep the default workflow lean\n\n- Default to direct T1 work in code and tests. Do not create OpenSpec artifacts from task wording alone.\n- OpenSpec is explicit opt-in: use `--tier T2` for a change workspace or `--tier T3` for change plus plan only when the user requests OpenSpec or the caller explicitly selects T2\u002FT3.\n- Treat the commit, targeted test output, and compact final handoff as the routine task record.\n\n## Agent and subagent collaboration\n\nBefore parallel work, inspect the field with the MCP `list_agents` tool or the CLI fallback:\n\n```sh\ngx mcp list-agents --no-prs\n```\n\nFor every file a parallel writer may touch, check ownership, create an isolated lane, then claim it:\n\n```sh\ngx mcp who-owns path\u002Fto\u002Ffile.ts\ngx branch start --new --no-transfer \"\u003Cbounded task>\" \"\u003Cagent>\"\ngx locks claim --branch \"\u003Cagent-branch>\" path\u002Fto\u002Ffile.ts\n```\n\n- **One owner per file.** Split writer subagents by non-overlapping files or modules. Read-only scouts may share a surface; writers may not.\n- Give each subagent one bounded objective and a verification command. Do not ask one subagent to explore, redesign, implement, and review the same slice.\n- Re-check `list_agents` before expanding scope. If another lane owns or is already editing a file, coordinate a handoff or choose a different slice.\n- Return a compact handoff: conclusion, `file:line`, changed files, verification result, and blocker or next action. Do not relay full logs.\n- Finish each writing lane through its own PR once. Never run duplicate finish commands for the same branch.\n\n## Optional cross-machine Git lock and radar state\n\nLocal worktree locks remain the default. When agents on multiple machines use the same writable Git remote, opt into shared state per repository:\n\n```sh\ngx locks shared-enable --remote origin\ngx locks shared-status\n```\n\nAfter enablement, ordinary `claim`, `allow-delete`, `validate`, `release`, and `status` commands also use atomic remote refs under `refs\u002Fgitguardex\u002Flocks\u002F*`. `gx mcp list-agents` combines local worktrees with remote `agent\u002F*` branches and shared lock owners; `gx mcp who-owns` resolves the same shared owner data.\n\n- Shared mode is explicit and fail-closed: if the remote is unreachable, rejects custom refs, or returns invalid metadata, do not keep a local-only claim or approve a commit.\n- The remote must allow authenticated custom-ref creation, compare-and-swap updates, and deletion. GitGuardex stores validated JSON metadata in commits; it never executes remote metadata, and publishes only a hashed machine identifier.\n- Push the agent branch when full remote-lane visibility matters. A lock still makes an unpushed branch visible as a lock-owning remote lane.\n- Release locks before disabling shared state or abandoning a lane. Cross-machine locks are not automatically reaped because one host cannot safely prove that another host's process is dead.\n- `gx locks shared-disable` stops shared reads and writes for this repository; it does not delete remote refs owned by any machine.\n\n## Finish checklist in Codex\n\nBefore starting a long `gx finish`, `gx ship`, or `gx branch finish`, show this compact Markdown checklist in commentary and update it from the CLI's `[gx:finish]` transition lines:\n\n- [ ] Prepare branch\n- [ ] Local preflight\n- [ ] Push and open PR\n- [ ] AI review\n- [ ] Review autofix\n- [ ] CI checks\n- [ ] Merge\n- [ ] Cleanup\n\nRender each state-change update as a compact card, not as a raw log dump:\n\n### ? GX Finish · `3\u002F8`\n`agent\u002F...` → `main` · ? **AI review** · `00:44`\n\n- ✅ Prepare branch\n- ✅ Local preflight\n- ✅ Push and open PR · `#123`\n- ? AI review\n- ⬜ Review autofix\n- ⬜ CI checks\n- ⬜ Merge\n- ⬜ Cleanup\n\nStart the command with a short initial yield, then poll the background process every 15-30 seconds. Between polls, post an updated checklist when the active stage changes; during a quiet long stage, name the active stage and elapsed time at most every 30 seconds. Never leave the user with only Codex's generic **Working** or **Waiting for background terminal** label. Do not launch a duplicate finish while the first process is still alive, and do not mark a checkbox complete until a matching `✓` line or equivalent terminal evidence arrives.\n\nTreat `? Cleanup` as finished-but-best-effort, not as unconditional success: surface any cleanup warnings beside it instead of converting it to `✅`.\n\nStructured phase events are persisted privately under `.omx\u002Fstate\u002Ffinish-runs\u002F*.jsonl` (`0700` directory, `0600` files). Prefer that JSONL stream over parsing narrative logs when exact state is needed; the visual output prints the active event-file path.\n\nFor agent transcripts, opt into `gx branch finish --agent-quiet` (or `gx finish --agent-quiet`)\nalongside the same required finish flags and gates. It prints one run ID\u002Fevent-file pointer,\nretains warnings and blockers, and ends with a bounded summary; it does not skip verification.\nRead only new observations with `gx finish events --run \u003Crun-id> --wait-ms 30000`, then\npass the returned `--cursor \u003Ccursor>` on the next read. Use the same repository target as\nthe finish invocation. Let this bounded CLI wait do the polling instead of repeatedly\nasking the model to reread the whole log. Events are observations, never merge authority.\n\nOpt-in bounded observations (defaults remain unchanged):\n- MCP `my_context`: pass `max_bytes` to bound UTF-8 JSON tool text. Ownership,\n  conflicts and safety fields are mandatory; `budget-too-small` requires a larger\n  budget or smaller file batch. Check `complete`\u002F`omitted`; fetch full peers only\n  when needed. This budgets bytes, not model tokens.\n- `gx watch --pr-cache-ms 5000` shares PR probes with a short in-process TTL keyed\n  by repository, remote, branch and HEAD. Observation age is shown; dirty state\n  stays fresh. The cache is never used by merge gates.\n- `gx finish events --run \u003Cid> --view state`: merge `stages` patches by stage,\n  retain every `alerts` record, and resume with the opaque cursor. Paginated\n  snapshots are incomplete until `snapshotComplete: true`. Raw view remains the\n  default; reset explicitly on invalid\u002Freplaced\u002Ftruncated cursors.\n- `gx agents set-status --activity working --dedupe-surface`: skip identical\n  successful pane-label writes, but still renew session activity on every call.\n  Changed targets and failed writes retry. This never infers agent completion.\n\nFor setup guidance, use `gx prompt --list-parts` then `gx prompt --part \u003Cname>` for the\nneeded slice; use `--exec --part \u003Cname>` only for command-capable slices. Do not repeatedly\nload the full setup prompt or remove the mandatory safety contract to save tokens.\n\nWhen inspecting or verifying, prefer `rtk` compact wrappers if available (`rtk git status`, `rtk grep`, `rtk test \u003Ccmd>`, and noisy gx reads like `rtk gx status` \u002F `rtk gx doctor`). Do not wrap commands whose stdout is parsed by scripts (`--json`, `--porcelain`, exact stdout contracts) or shell-ready output (`gx prompt --exec`).\n\nTo shrink gx's own large narrative output (e.g. `gx prompt`, `gx prompt --snippet`) before it lands in your context, set `GUARDEX_COMPRESS_CMD=\"\u003Cstdin->stdout filter>\"`; gx routes that output through the filter (terse\u002Fnon-TTY mode, fail-open, JSON skipped). Unset = byte-for-byte unchanged. Confirm it is wired with `gx status` or `gx doctor` — both print a `Token compression` line and flag a configured-but-missing binary (doctor's warning is advisory and never changes its safe\u002Funsafe exit code).\n",{"name":5,"description":7},"main","7546eb50cf658dad033c5a047a7ec89c9d1ee7f4","templates\u002Fcodex\u002Fskills\u002Fgitguardex",4,"https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex","https:\u002F\u002Fgitguardex.com\u002F","https:\u002F\u002Fgithub.com\u002Fopencue","Building in the open. From AI-powered workflows and developer tools to web experiences, automation, and e-commerce — practical software for real-world problems.","2026-11-12 18:11:01",{"source":8,"skillName":5,"commands":54,"prompt":64,"zipAvailable":65},[55,58,61],{"tool":56,"command":57},"npx","npx skills add opencue\u002Fgitguardex --skill gitguardex",{"tool":59,"command":60},"bunx","bunx skills add opencue\u002Fgitguardex --skill gitguardex",{"tool":62,"command":63},"pnpm","pnpm dlx skills add opencue\u002Fgitguardex --skill gitguardex","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Ftemplates\u002Fcodex\u002Fskills\u002Fgitguardex\u002FSKILL.md",true,[],[68,86,108,129,145,151],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":39},"01M6QEBH0VSDT0T2HX5XE3VW5F","opensrc","Fetch exact dependency or external repository source before inspecting internals. Use for package implementation, upstream behavior, or external GitHub source lookup.","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Ftemplates\u002Fcodex\u002Fskills\u002Fopensrc\u002FSKILL.md",[],[70,75,76,38,77,78,79,34,80,81,82,83,84,85],"Bash","source","upstream","implementation","dependency","fetch","behavior","lookup","repository","package","github",{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":95,"publishedAt":39},"01M6QEBDNSWK5R390QHGFWP601","gx-act","Run GitHub Actions workflows locally with nektos\u002Fact before pushing, so CI failures are caught on the laptop and the PR can be squash-merged on the first remote run.","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Ftemplates\u002Fcodex\u002Fskills\u002Fgx-act\u002FSKILL.md",[92],{"categoryNo":93,"categoryName":94,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署",[96,97,98,99,100,101,102,103,25,104,85,105,106,107],"gx act","act","squash","first","github actions","ci","actions","run","workflows","can","with","remote",{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":115,"publishedAt":39},"01M6QEBAQ8HA4ZGNDS03AB25Y7","guardex-merge-skills-to-dev","Use when you need to merge SKILL.md updates from agent branches\u002Fworktrees into the local base branch (default: dev) with the multiagent-safety flow.","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Ftemplates\u002Fcodex\u002Fskills\u002Fguardex-merge-skills-to-dev\u002FSKILL.md",[114],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[116,117,118,119,120,121,106,122,123,124,125,126,34,33,127,29,128],"guardex","guardex merge skills to dev","dev","merge","updates","you","local","branches","base","default","worktrees","multiagent","flow",{"skillNo":130,"name":131,"nameZh":6,"description":132,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":133,"categories":134,"keywords":135,"publishedAt":39},"01M6QEAYMVDN8RZCVHQS7YNCBJ","pr-lens","Generate local PR Lens architecture and data-flow diagrams alongside a GitGuardex Codex or Claude PR review.","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Fskills\u002Fpr-lens\u002FSKILL.md",[],[136,137,25,138,139,24,140,5,122,141,142,143,128,144],"pr lens","lens","architecture","codex","pr review","diagrams","data-flow","data","generate",{"skillNo":146,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":147,"categories":148,"keywords":150,"publishedAt":39},"01M6QEAVDZQRKP05NHQ2AWR3MY","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Fskills\u002Fgx-act\u002FSKILL.md",[149],{"categoryNo":93,"categoryName":94,"taxonomyType":21,"taxonomyTypeDesc":22},[96,97,99,105,106,107,98,100,101,102,103,25,104,85],{"skillNo":152,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":153,"categories":154,"keywords":156,"publishedAt":39},"01M6QEARBHH7N7VTWS69P2E9MW","https:\u002F\u002Fgithub.com\u002Fopencue\u002Fgitguardex\u002Fblob\u002Fmain\u002Fskills\u002Fguardex-merge-skills-to-dev\u002FSKILL.md",[155],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[118,116,117,119,127,29,33,128,122,106,121,125,34,126,120,124,123]]