[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M753G2291VZ7AY3BETFS8EA2":3,"skill-same:01M753G2291VZ7AY3BETFS8EA2":74},{"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":28,"commitSha":29,"skillDirectory":30,"forksCount":14,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":33,"ownerDescription":34,"updatedAt":35,"install":36,"resources":49},"01M753G2291VZ7AY3BETFS8EA2","cmux-orchestration",null,"Control cmux panes and surfaces to coordinate terminal agent workers. Use when a top-level agent needs to launch, prompt, monitor, or manage Claude Opus and Codex GPT-5.5 TUIs in the current cmux workspace, run plan-first delegations (Claude Code plan mode, Codex Plan mode), track worker state, delegate bounded tasks, or verify worker reports.","haryelramalho\u002Fskills","haryelramalho","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F38708454?v=4","01M753ESC6XNS6GXGV5X9YMKGA","Haryel Ramalho","01M753ESCSPRNDTEJRVRYY1VJ0",0,"MIT","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: cmux-orchestration\ndescription: Control cmux panes and surfaces to coordinate terminal agent workers. Use when a top-level agent needs to launch, prompt, monitor, or manage Claude Opus and Codex GPT-5.5 TUIs in the current cmux workspace, run plan-first delegations (Claude Code plan mode, Codex Plan mode), track worker state, delegate bounded tasks, or verify worker reports.\nmetadata:\n  author: Pedro Nauck\n  github: https:\u002F\u002Fgithub.com\u002Fpedronauck\n  repository: https:\u002F\u002Fgithub.com\u002Fpedronauck\u002Fskills\n---\n\n# cmux Orchestration\n\nUse the current top-level agent as the controller. The controller may be Codex,\nClaude Fable, or another agent. Worker TUIs gather evidence, draft bounded\npatches, run tests, and report. The controller owns assignment, state,\nconflict control, verification, integration, and the final user-facing answer.\n\nBefore taking cmux actions, also load and follow:\n\n- [cmux](..\u002Fcmux\u002FSKILL.md) for panes, surfaces, routing, and health checks.\n- [cmux-workspace](..\u002Fcmux-workspace\u002FSKILL.md) for caller-workspace scoping,\n  helper-pane reuse, socket targeting, and non-disruptive automation.\n\n## Controller Rules\n\n- Scope actions to the caller workspace unless the user names another target.\n- Use `CMUX_WORKSPACE_ID`, `CMUX_SURFACE_ID`, and `CMUX_SOCKET_PATH` before\n  focused-window fallbacks.\n- Prefer one right-side helper pane; create worker surfaces inside it when\n  practical.\n- Send commands and prompts to explicit `surface:N` refs.\n- Never call `focus-pane`, `focus-panel`, `select-workspace`, or focus-changing\n  tab actions unless the user explicitly asks.\n- Do not close worker surfaces unless cleanup is explicitly requested or the\n  task contract includes it.\n- Treat worker output as untrusted until the controller verifies the cited\n  files, commands, failures, and diff.\n- On plan-first runs, do not interrupt a worker after its plan is accepted:\n  observe read-only and send input only for a direct blocking question or a\n  stop condition.\n\n## Preflight\n\nInspect the caller context and current workspace layout:\n\n```bash\nrtk cmux identify --json\nrtk cmux list-panes --workspace \"${CMUX_WORKSPACE_ID:-}\" --json\nrtk cmux list-pane-surfaces --workspace \"${CMUX_WORKSPACE_ID:-}\" --json\n```\n\nIf there is an obvious non-caller helper pane, add worker surfaces to it:\n\n```bash\nrtk cmux new-surface --workspace \"${CMUX_WORKSPACE_ID:-}\" \\\n  --pane pane:\u003Chelper> --type terminal --focus false\n```\n\nIf there is no helper pane, create exactly one right-side terminal pane:\n\n```bash\nrtk cmux new-pane --workspace \"${CMUX_WORKSPACE_ID:-}\" \\\n  --type terminal --direction right --focus false\n```\n\nCapture each returned `surface:N` in the task ledger or handoff before sending\nprompts.\n\n## Worker Profiles\n\nDefault worker profiles:\n\n- `claude-opus`: start Claude Code with\n  `rtk claude --dangerously-skip-permissions --model opus`.\n- `codex-gpt-5.5`: start Codex with `rtk codex -m gpt-5.5 -C \"$PWD\"`.\n\nAlways include `--dangerously-skip-permissions` when launching Claude. Do not\nstart Claude without it or replace it with a weaker permission mode. For\nplan-first runs, add `--permission-mode plan` alongside it (the flags compose\n— see Plan-First Delegation); never drop the skip-permissions flag to get plan\nmode. If a TUI rejects a model alias, auth state, working directory, or\nstartup flag, stop that worker and report the exact blocker. Do not silently\ndowngrade models or switch tools without user approval.\n\nLaunch workers by explicit surface:\n\n```bash\nrtk cmux send --surface surface:\u003Cclaude> \\\n  \"rtk claude --dangerously-skip-permissions --model opus --name cmux-claude-opus\\n\"\nrtk cmux send --surface surface:\u003Ccodex> \\\n  \"rtk codex -m gpt-5.5 -C \\\"$PWD\\\"\\n\"\n```\n\n## Plan-First Delegation (Plan Mode)\n\nFor slices that need investigation and a reviewed plan before edits\n(root-cause fixes, multi-file changes, unfamiliar code), run the worker\nplan-first: launch in plan mode → submit the packet → watch and answer\nclarifying questions → review the plan → accept → hands-off until done.\n\n**STOP. Read [references\u002Fplan-mode.md](references\u002Fplan-mode.md) in full before\nlaunching a worker in plan mode.** The launch flags, shift+tab sequences,\nstatus-line checks, and acceptance menus are exact and differ per TUI; this\nsection is a tripwire, not the procedure.\n\nSummary of the per-TUI entry points (the reference has the full flow):\n\n- Claude Code: launch with\n  `rtk claude --dangerously-skip-permissions --permission-mode plan --model opus`\n  (the flags compose; the packet may ride the launch line as a CLI argument).\n  Accept via the `Would you like to proceed?` menu — pick the auto-accept\n  option so the worker runs unattended.\n- Codex: no plan-mode CLI flag. Launch bare (never with the packet as a CLI\n  argument), press shift+tab until the status line reads `Plan mode`, then\n  submit the packet. Accept via `Implement this plan?` → option 1 (keeps the\n  investigation context).\n\nPlan mode is interactive until acceptance: answer the worker's clarifying\nquestions with the packet context. After acceptance, do not interrupt —\nobserve read-only until the final completion report, then verify claims per\nMonitoring and Verification.\n\n## Prompt Submission to Running TUIs\n\n`cmux send \"...\\n\"` only submits when the target surface is a plain shell\n(like the launch lines above). Inside a running TUI (Claude Code, Codex), the\ntrailing `\\n` is treated as pasted text or a soft newline — the prompt sits in\nthe input box unsubmitted and the worker stalls silently.\n\nRules:\n\n- Prefer passing the initial prompt as a CLI argument at launch\n  (`rtk claude --dangerously-skip-permissions \"…prompt…\"`,\n  `rtk codex \"…prompt…\"`). This avoids the TUI-ready race and the submission\n  problem entirely. Exception: Codex plan-first runs must launch bare — see\n  Plan-First Delegation.\n- For any follow-up prompt to an already-running TUI, always send the text and\n  then submit with an explicit Enter:\n\n  ```bash\n  rtk cmux send --surface surface:\u003Cworker> \"…follow-up prompt…\"\n  rtk cmux send-key --surface surface:\u003Cworker> enter\n  ```\n\n- Confirm submission with `rtk cmux read-screen --surface surface:\u003Cworker>`:\n  the input line must be empty and the TUI spinner\u002Fworking indicator active.\n  Treat an unconfirmed send as not delivered.\n\n## Delegation Packets\n\nSend worker prompts as standalone contracts. Include:\n\n- repo path and exact objective\n- worker role and model\n- execution mode: plan-first (plan mode) or direct\n- files, packages, or surfaces in scope\n- files and behaviors explicitly out of scope\n- claimed files or work slice to avoid conflicts\n- expected evidence format: files, line refs, commands, diffs, failures,\n  screenshots, and uncertainty\n- verification commands or browser flows to run\n- stop conditions: unexpected code shape, repeated command failure, auth\u002Fmodel\n  blocker, or need for out-of-scope edits\n\nPrefer small, independent worker slices. Do not assign overlapping file claims\nunless the controller plans to integrate and resolve conflicts immediately.\n\n## State Registry\n\nMaintain a compact registry in the task ledger or handoff:\n\n- controller identity\n- worker role and model\n- `workspace:N`, `pane:N`, and `surface:N`\n- prompt\u002Fobjective sent and start time\n- current status: starting, planning, plan-review, plan-accepted, running,\n  waiting, blocked, reported, verified\n- claimed files or work slice\n- worker-reported commands and results\n- controller verification performed\n- final disposition: accepted, rejected, superseded, or blocked\n\nFor long runs, attach progress to the caller workspace with cmux sidebar state:\n\n```bash\nrtk cmux set-status orchestration \"running\" \\\n  --workspace \"${CMUX_WORKSPACE_ID:-}\" --color \"#ff9500\"\nrtk cmux set-progress 0.4 --label \"Workers running\" \\\n  --workspace \"${CMUX_WORKSPACE_ID:-}\"\nrtk cmux log --workspace \"${CMUX_WORKSPACE_ID:-}\" --level info -- \\\n  \"Started cmux orchestration\"\n```\n\nClear status\u002Fprogress when orchestration finishes or is abandoned.\n\n## Monitoring and Verification\n\n- Poll or inspect worker surfaces without changing focus.\n- Ask a worker for a concise status summary when output is unclear.\n- Verify high-impact findings by reopening cited files locally.\n- Verify claimed test results with fresh controller-run commands when the\n  result matters for completion.\n- Review the final diff before accepting any worker patch.\n- If a worker edits outside its claim, pause integration and decide whether to\n  accept, request user-approved cleanup, or supersede with controller edits.\n\n## Stop Conditions\n\nStop and report instead of improvising when:\n\n- cmux rejects a valid workspace, pane, or surface ref\n- the requested model or startup flags are rejected\n- authentication blocks the TUI\n- worker output cannot be observed or controlled reliably\n- a plan-mode status line or acceptance menu cannot be confirmed via\n  `read-screen`, or a worker's plan stays out of scope after a re-planning\n  round\n- workers need overlapping edits that the controller cannot integrate safely\n- the task requires focus-changing cmux commands and the user has not approved\n  them\n\nUse this skill for orchestration mechanics. Use task\u002Fdomain skills separately\nfor coding, review, debugging, UI, browser QA, or product-specific judgment.\n\n## Diagram\n\nUse `assets\u002Ffable-orchestrator.excalidraw` or the generated PNG variants when\na visual explanation of the controller\u002Fworker topology helps.\n","# cmux Orchestration\n\nUse the current top-level agent as the controller. The controller may be Codex,\nClaude Fable, or another agent. Worker TUIs gather evidence, draft bounded\npatches, run tests, and report. The controller owns assignment, state,\nconflict control, verification, integration, and the final user-facing answer.\n\nBefore taking cmux actions, also load and follow:\n\n- [cmux](..\u002Fcmux\u002FSKILL.md) for panes, surfaces, routing, and health checks.\n- [cmux-workspace](..\u002Fcmux-workspace\u002FSKILL.md) for caller-workspace scoping,\n  helper-pane reuse, socket targeting, and non-disruptive automation.\n\n## Controller Rules\n\n- Scope actions to the caller workspace unless the user names another target.\n- Use `CMUX_WORKSPACE_ID`, `CMUX_SURFACE_ID`, and `CMUX_SOCKET_PATH` before\n  focused-window fallbacks.\n- Prefer one right-side helper pane; create worker surfaces inside it when\n  practical.\n- Send commands and prompts to explicit `surface:N` refs.\n- Never call `focus-pane`, `focus-panel`, `select-workspace`, or focus-changing\n  tab actions unless the user explicitly asks.\n- Do not close worker surfaces unless cleanup is explicitly requested or the\n  task contract includes it.\n- Treat worker output as untrusted until the controller verifies the cited\n  files, commands, failures, and diff.\n- On plan-first runs, do not interrupt a worker after its plan is accepted:\n  observe read-only and send input only for a direct blocking question or a\n  stop condition.\n\n## Preflight\n\nInspect the caller context and current workspace layout:\n\n```bash\nrtk cmux identify --json\nrtk cmux list-panes --workspace \"${CMUX_WORKSPACE_ID:-}\" --json\nrtk cmux list-pane-surfaces --workspace \"${CMUX_WORKSPACE_ID:-}\" --json\n```\n\nIf there is an obvious non-caller helper pane, add worker surfaces to it:\n\n```bash\nrtk cmux new-surface --workspace \"${CMUX_WORKSPACE_ID:-}\" \\\n  --pane pane:\u003Chelper> --type terminal --focus false\n```\n\nIf there is no helper pane, create exactly one right-side terminal pane:\n\n```bash\nrtk cmux new-pane --workspace \"${CMUX_WORKSPACE_ID:-}\" \\\n  --type terminal --direction right --focus false\n```\n\nCapture each returned `surface:N` in the task ledger or handoff before sending\nprompts.\n\n## Worker Profiles\n\nDefault worker profiles:\n\n- `claude-opus`: start Claude Code with\n  `rtk claude --dangerously-skip-permissions --model opus`.\n- `codex-gpt-5.5`: start Codex with `rtk codex -m gpt-5.5 -C \"$PWD\"`.\n\nAlways include `--dangerously-skip-permissions` when launching Claude. Do not\nstart Claude without it or replace it with a weaker permission mode. For\nplan-first runs, add `--permission-mode plan` alongside it (the flags compose\n— see Plan-First Delegation); never drop the skip-permissions flag to get plan\nmode. If a TUI rejects a model alias, auth state, working directory, or\nstartup flag, stop that worker and report the exact blocker. Do not silently\ndowngrade models or switch tools without user approval.\n\nLaunch workers by explicit surface:\n\n```bash\nrtk cmux send --surface surface:\u003Cclaude> \\\n  \"rtk claude --dangerously-skip-permissions --model opus --name cmux-claude-opus\\n\"\nrtk cmux send --surface surface:\u003Ccodex> \\\n  \"rtk codex -m gpt-5.5 -C \\\"$PWD\\\"\\n\"\n```\n\n## Plan-First Delegation (Plan Mode)\n\nFor slices that need investigation and a reviewed plan before edits\n(root-cause fixes, multi-file changes, unfamiliar code), run the worker\nplan-first: launch in plan mode → submit the packet → watch and answer\nclarifying questions → review the plan → accept → hands-off until done.\n\n**STOP. Read [references\u002Fplan-mode.md](references\u002Fplan-mode.md) in full before\nlaunching a worker in plan mode.** The launch flags, shift+tab sequences,\nstatus-line checks, and acceptance menus are exact and differ per TUI; this\nsection is a tripwire, not the procedure.\n\nSummary of the per-TUI entry points (the reference has the full flow):\n\n- Claude Code: launch with\n  `rtk claude --dangerously-skip-permissions --permission-mode plan --model opus`\n  (the flags compose; the packet may ride the launch line as a CLI argument).\n  Accept via the `Would you like to proceed?` menu — pick the auto-accept\n  option so the worker runs unattended.\n- Codex: no plan-mode CLI flag. Launch bare (never with the packet as a CLI\n  argument), press shift+tab until the status line reads `Plan mode`, then\n  submit the packet. Accept via `Implement this plan?` → option 1 (keeps the\n  investigation context).\n\nPlan mode is interactive until acceptance: answer the worker's clarifying\nquestions with the packet context. After acceptance, do not interrupt —\nobserve read-only until the final completion report, then verify claims per\nMonitoring and Verification.\n\n## Prompt Submission to Running TUIs\n\n`cmux send \"...\\n\"` only submits when the target surface is a plain shell\n(like the launch lines above). Inside a running TUI (Claude Code, Codex), the\ntrailing `\\n` is treated as pasted text or a soft newline — the prompt sits in\nthe input box unsubmitted and the worker stalls silently.\n\nRules:\n\n- Prefer passing the initial prompt as a CLI argument at launch\n  (`rtk claude --dangerously-skip-permissions \"…prompt…\"`,\n  `rtk codex \"…prompt…\"`). This avoids the TUI-ready race and the submission\n  problem entirely. Exception: Codex plan-first runs must launch bare — see\n  Plan-First Delegation.\n- For any follow-up prompt to an already-running TUI, always send the text and\n  then submit with an explicit Enter:\n\n  ```bash\n  rtk cmux send --surface surface:\u003Cworker> \"…follow-up prompt…\"\n  rtk cmux send-key --surface surface:\u003Cworker> enter\n  ```\n\n- Confirm submission with `rtk cmux read-screen --surface surface:\u003Cworker>`:\n  the input line must be empty and the TUI spinner\u002Fworking indicator active.\n  Treat an unconfirmed send as not delivered.\n\n## Delegation Packets\n\nSend worker prompts as standalone contracts. Include:\n\n- repo path and exact objective\n- worker role and model\n- execution mode: plan-first (plan mode) or direct\n- files, packages, or surfaces in scope\n- files and behaviors explicitly out of scope\n- claimed files or work slice to avoid conflicts\n- expected evidence format: files, line refs, commands, diffs, failures,\n  screenshots, and uncertainty\n- verification commands or browser flows to run\n- stop conditions: unexpected code shape, repeated command failure, auth\u002Fmodel\n  blocker, or need for out-of-scope edits\n\nPrefer small, independent worker slices. Do not assign overlapping file claims\nunless the controller plans to integrate and resolve conflicts immediately.\n\n## State Registry\n\nMaintain a compact registry in the task ledger or handoff:\n\n- controller identity\n- worker role and model\n- `workspace:N`, `pane:N`, and `surface:N`\n- prompt\u002Fobjective sent and start time\n- current status: starting, planning, plan-review, plan-accepted, running,\n  waiting, blocked, reported, verified\n- claimed files or work slice\n- worker-reported commands and results\n- controller verification performed\n- final disposition: accepted, rejected, superseded, or blocked\n\nFor long runs, attach progress to the caller workspace with cmux sidebar state:\n\n```bash\nrtk cmux set-status orchestration \"running\" \\\n  --workspace \"${CMUX_WORKSPACE_ID:-}\" --color \"#ff9500\"\nrtk cmux set-progress 0.4 --label \"Workers running\" \\\n  --workspace \"${CMUX_WORKSPACE_ID:-}\"\nrtk cmux log --workspace \"${CMUX_WORKSPACE_ID:-}\" --level info -- \\\n  \"Started cmux orchestration\"\n```\n\nClear status\u002Fprogress when orchestration finishes or is abandoned.\n\n## Monitoring and Verification\n\n- Poll or inspect worker surfaces without changing focus.\n- Ask a worker for a concise status summary when output is unclear.\n- Verify high-impact findings by reopening cited files locally.\n- Verify claimed test results with fresh controller-run commands when the\n  result matters for completion.\n- Review the final diff before accepting any worker patch.\n- If a worker edits outside its claim, pause integration and decide whether to\n  accept, request user-approved cleanup, or supersede with controller edits.\n\n## Stop Conditions\n\nStop and report instead of improvising when:\n\n- cmux rejects a valid workspace, pane, or surface ref\n- the requested model or startup flags are rejected\n- authentication blocks the TUI\n- worker output cannot be observed or controlled reliably\n- a plan-mode status line or acceptance menu cannot be confirmed via\n  `read-screen`, or a worker's plan stays out of scope after a re-planning\n  round\n- workers need overlapping edits that the controller cannot integrate safely\n- the task requires focus-changing cmux commands and the user has not approved\n  them\n\nUse this skill for orchestration mechanics. Use task\u002Fdomain skills separately\nfor coding, review, debugging, UI, browser QA, or product-specific judgment.\n\n## Diagram\n\nUse `assets\u002Ffable-orchestrator.excalidraw` or the generated PNG variants when\na visual explanation of the controller\u002Fworker topology helps.\n",{"name":5,"metadata":24,"description":7},{"author":25,"github":26,"repository":27},"Pedro Nauck","https:\u002F\u002Fgithub.com\u002Fpedronauck","https:\u002F\u002Fgithub.com\u002Fpedronauck\u002Fskills","main","a5b533be2d244bbf3e5606e851742f0e0ba99423","skills\u002Fcommunity\u002Fcmux-orchestration","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills","https:\u002F\u002Fgithub.com\u002Fharyelramalho","linkedin.com\u002Fin\u002Fharyelramalho\u002F","Software Engineer @ SKORR •  Former Vice President of the Junior Computer Association - TecnoJr","2026-07-16 01:34:58",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add haryelramalho\u002Fskills --skill cmux-orchestration",{"tool":42,"command":43},"bunx","bunx skills add haryelramalho\u002Fskills --skill cmux-orchestration",{"tool":45,"command":46},"pnpm","pnpm dlx skills add haryelramalho\u002Fskills --skill cmux-orchestration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002FSKILL.md",true,[50,56,62,66,70],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"agents\u002Fopenai.yaml","REFERENCE","参考文档",218,"https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002Fagents\u002Fopenai.yaml",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"assets\u002Ffable-orchestrator-dark.png","ASSET","资产文件",111350,"https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002Fassets\u002Ffable-orchestrator-dark.png",{"path":63,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":64,"sourceHtmlUrl":65},"assets\u002Ffable-orchestrator.excalidraw",19562,"https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002Fassets\u002Ffable-orchestrator.excalidraw",{"path":67,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":68,"sourceHtmlUrl":69},"assets\u002Ffable-orchestrator.png",111227,"https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002Fassets\u002Ffable-orchestrator.png",{"path":71,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":72,"sourceHtmlUrl":73},"references\u002Fplan-mode.md",7227,"https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcommunity\u002Fcmux-orchestration\u002Freferences\u002Fplan-mode.md",[75,82,89,96,103,110],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M753YDMKSDB7320THEQ2CVW5","to-prompt-opus","Use when preparing or rewriting prompts for Claude Opus 4.8, Anthropic API agents, XML-structured prompts, long-context document work, coding agents, research synthesis, frontend\u002Fdesign tasks, or model-specific prompt optimization. Do not use for GPT\u002FOpenAI prompts or raw context packaging without Opus-specific tuning.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fmine\u002Fto-prompt-opus\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M753YAP95W4K2MCEAEEAV6YA","to-prompt-gpt","Use when preparing or rewriting prompts for GPT-5.5, OpenAI API agents, Responses workflows, coding agents, research synthesis, support assistants, or model-specific prompt optimization. Do not use for Claude\u002FOpus prompts or raw context packaging without GPT-specific tuning.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fmine\u002Fto-prompt-gpt\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M753Y700AZ0GVJDNW9Q1Q54F","plan-refactor-qa","Plans and runs qa-report before executing Compozy refactor tasks. Use when the user provides a refactor or issue slug and wants Codex to inspect the matching .compozy\u002Ftasks directory, related PRD\u002FTechSpec\u002FADR\u002Ftask artifacts, affected code, and existing tests, then invoke $qa-report with a repository-grounded prompt. Do not use to run tests, execute qa-execution, implement code, or perform post-refactor verification.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fmine\u002Fplan-refactor-qa\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":19},"01M753XZENG0Z8S2ET75EM8CGB","multi-tool-branch-review","Orchestrates a multi-tool branch review of the current git branch against a base branch. By default it runs two independent thermo-nuclear maintainability audits, one on Codex GPT-5.5 xhigh and one on Claude Opus xhigh, then consolidates the findings into a single remediation plan via writing-plans. If the user explicitly asks for native reviews, it can add the Claude native review pass, the Codex native review pass, or both. Use when the user wants a deep cross-model review of a branch or PR before merge, or asks to review this branch with multiple tools or models. Encodes known infra failure modes and their fixes (missing Claude ACP adapter, Codex MCP auth hangs, false-positive report files, long jobs orphaned inside subagents). Do not use for reviewing a single uncommitted snippet, for non-git work, for running only one reviewer, or when the user only wants a quick lint or test run.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fmine\u002Fmulti-tool-branch-review\u002FSKILL.md",[],[],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"publishedAt":19},"01M753XQM40XCK55BH5WPA88EW","create-pr-with-template","Create a GitHub Pull Request with gh CLI and prefer repository PR templates automatically. Use this whenever the user asks to open\u002Fcreate a PR, especially when template compliance is needed. If multiple templates exist, ask the user which template to use.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fmine\u002Fcreate-pr-with-template\u002FSKILL.md",[],[],{"skillNo":111,"name":112,"nameZh":6,"description":113,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":114,"categories":115,"keywords":116,"publishedAt":19},"01M753XJMBJ8VFPRCDCC1JJ9G4","ui-craft","Provides guardrails for user-facing UI work: usability heuristics, accessibility floors, design-system discipline, component states, microcopy, motion, dark mode, responsive behavior, and human-AI UX. Use when designing, generating, reviewing, or refactoring visible product surfaces such as components, pages, dashboards, forms, dialogs, loading\u002Fempty\u002Ferror states, or AI interfaces. Do not use for backend-only work, infrastructure, CLI\u002FTUI design, or pure documentation editing.","https:\u002F\u002Fgithub.com\u002Fharyelramalho\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fcurated\u002Fui-craft\u002FSKILL.md",[],[]]