[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6D2W8PR40H3V9TBC409W0B5":3,"skill-same:01M6D2W8PR40H3V9TBC409W0B5":82},{"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":35,"linkOnly":36,"markdownContent":37,"markdownBody":38,"frontMatter":39,"refName":44,"commitSha":45,"skillDirectory":46,"forksCount":47,"repositoryHtmlUrl":48,"repositoryHomepageUrl":49,"ownerHtmlUrl":50,"ownerBlogUrl":51,"ownerDescription":52,"updatedAt":53,"install":54,"resources":67},"01M6D2W8PR40H3V9TBC409W0B5","delegate",null,"Live in-session delegation from Claude to Codex via the approval\u002Fdelegation packet contract","GeorgeQLe\u002Fagentic-skills","GeorgeQLe","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F21134126?v=4","01M6D2JP418S0VGJGF8M93EC3H","George Le","01M6D2JPBAKYD82TA5FS6HKQ3D",1,"MIT","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Fpacks\u002Fagent-bridge\u002Fclaude\u002Fdelegate\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000037","律师与法务","PROFESSION","适用职业",[5,24,25,26,27,28,29,30,31,32,33,34],"codex","shipping","delegation","page","v0","session","approval","alignment","fallback","contract","live","2026-11-10 10:12:30",false,"---\nname: delegate\ndescription: Live in-session delegation from Claude to Codex via the approval\u002Fdelegation packet contract\ntype: shipping\nversion: v0.1\nrequired_conventions: [alignment-page]\nargument-hint: \"[target-skill] [--allow-dirty \u003Cglob>] [--inline-fallback]\"\n---\n\n# Delegate\n\nDelegate the next step's execution to Codex **live, inside the current Claude session**. Produces an `approved` packet via the shared contract (same helpers `\u002Fhandoff --target=codex` uses), then invokes Codex synchronously with `codex exec \"\u003Ctarget-skill> --execute-approved\"`. Claude stays the orchestrator; Codex executes; Claude resumes once Codex returns.\n\nUnlike `\u002Fhandoff --target=codex` (async — user resumes later), `\u002Fdelegate` is live cross-CLI transport. The target skill defaults to `$exec`; other executor skills (e.g. `$ship`) may be named but `$exec` is the primary use case.\n\n## Mode requirement\n\n`\u002Fdelegate` is **`hybrid`-only by design**. `claude-only` has no executor to delegate to; `codex-only` plans in Codex directly. In either non-hybrid mode, the skill exits non-zero with a `mode-mismatch:` reason and touches no packet state.\n\n## Transport dependency\n\n`\u002Fdelegate` requires the `codex` binary on `PATH`. If missing (or auth-failed, or the start marker never prints), the skill falls cleanly into the **pre-start-failure** branch of the fallback matrix — it does not crash, and the packet stays at `approved`.\n\n## Process\n\n1. **Resolve agent mode** via `.\u002Fscripts\u002Fagent-mode.sh`. If the effective mode is anything other than `hybrid`, exit non-zero with a `mode-mismatch:` message naming the resolved mode. Do not touch the packet.\n\n2. **Parse arguments** from `$ARGUMENTS`:\n   - First positional token (if any) is the target skill; defaults to `$exec`.\n   - Collect repeatable `--allow-dirty \u003Cglob>` flags (shell-glob semantics, same as `scripts\u002Fapproved-plan.sh check`). Note: back-to-back hybrid cycles rewrite the `tasks\u002Fapproved-plan.md` mirror on each `consume`; commit it between cycles, or pass `--allow-dirty tasks\u002Fapproved-plan.md` so the next `draft` tolerates the uncommitted mirror.\n   - `--inline-fallback` auto-selects inline Claude execution in the pre-start-failure branch without prompting.\n\n3. **Derive `phase` \u002F `step` \u002F `title` from `tasks\u002Ftodo.md`**, using the same rules as `\u002Fhandoff --target=codex`:\n   - First unchecked `- [ ]` under the `### Active Step Plan` block.\n   - Fallback: first unchecked `- [ ]` under the current `## Phase N` header.\n   - Parse `Phase N` and `Step N.X` tokens from the surrounding context; the checkbox text (stripped of `- [ ] **Step N.X** — `) becomes `title`.\n\n4. **Produce the approval packet**:\n   1. `.\u002Fscripts\u002Fapproved-plan.sh draft --phase \"Phase N\" --step \"Step N.X\" --title \"\u003Ctitle>\"` plus any `--allow-dirty \u003Cglob>` flags. Surface the helper's single-line failure reason verbatim if it fails.\n   2. Pretty-print the drafted packet (`jq . .agents\u002Fapproved-plan.json`).\n   3. Ask exactly one concise question: *\"Approve this packet for Codex execution via \u002Fdelegate?\"*\n   4. On yes → `.\u002Fscripts\u002Fapproved-plan.sh approve`. On no → leave at `draft` and stop; the user can approve later or discard with `.\u002Fscripts\u002Fapproved-plan.sh supersede`.\n\n5. **Invoke Codex synchronously**:\n   - Record a start-marker timestamp (ISO-8601 UTC) before the exec call so the fallback matrix can classify outcomes.\n   - Shell out to `codex exec \"\u003Ctarget-skill> --execute-approved\"`. Capture stdout, stderr, and the exit code.\n   - The Step 4 consumer (`$exec --execute-approved`) is what actually runs `scripts\u002Fapproved-plan.sh check && consume` inside Codex. `\u002Fdelegate` is transport only.\n\n6. **Classify via the safe-fallback matrix — never blind-retry**:\n\n   - **Pre-start failure** — `codex` binary not found, auth failure, or the start marker was never printed by Codex before exit:\n     - Packet stays at `approved` (no writer ran).\n     - Offer two options: (a) run inline in Claude (same session, no CLI boundary), or (b) keep the packet `approved` for later manual `$exec --execute-approved` in Codex.\n     - If `--inline-fallback` was passed, auto-select (a) without prompting.\n     - Under an `agent-team` execution profile, inline fallback is a **downgrade** requiring explicit interactive confirmation unless `--inline-fallback` is already on the command line.\n\n   - **Success** — Codex exited 0, printed `Approved packet consumed: …`, and `tasks\u002Fapproved-plan.md` reports `lifecycle: consumed`:\n     - Step 4's `consume` already flipped `approved → consumed`.\n     - Read `tasks\u002Fapproved-plan.md` to confirm `lifecycle: consumed`. If confirmed, report success and hand back to the user.\n     - If the mirror does not report `consumed`, treat as the ambiguous case below.\n\n   - **Ambiguous** — non-zero exit, timeout, or crash after the start marker printed (Codex may have started mutating state):\n     - Call `.\u002Fscripts\u002Fapproved-plan.sh mark-uncertain`. The packet lifecycle becomes `uncertain`.\n     - Prompt the user with three explicit options: *inspect repo state manually \u002F discard the packet via `.\u002Fscripts\u002Fapproved-plan.sh supersede` \u002F continue inline in Claude from the current repo state*.\n     - **Never blind-retry `codex exec` on the same packet.** Re-executing against a packet whose consumer may have partially mutated state risks double-execution.\n\n7. **Report** the classified outcome, the final lifecycle, and whatever the user chose from the fallback prompt.\n\n## Constraints\n\n- **Hybrid-only.** Non-hybrid modes exit with `mode-mismatch:` before any packet operation.\n- **Never blind-retry cross-CLI.** The invariant. Ambiguous failures always go through `mark-uncertain` + explicit user prompt.\n- **Contract is frozen.** Do not edit `docs\u002Foperating-modes.md`, `specs\u002Fapproved-plan.schema.json`, or add new lifecycle states \u002F freshness checks. `uncertain` is already in the schema enum; Step 6 is just its first writer.\n- **`mark-uncertain` is only valid from `approved`.** The helper rejects `draft`, `consumed`, `stale`, `superseded`, and `uncertain` source states.\n- **`codex` binary is a transport dependency.** Fall cleanly into pre-start-failure if missing; do not crash.\n\n## Alignment Page\n\nFollow the shared alignment-page convention via the packaged convention resolver; output path is `alignment\u002Fdelegate-{topic}.html`.\n\n## Default Shipping Contract\n\nFollow the shared shipping contract convention in CLAUDE.md.\n","# Delegate\n\nDelegate the next step's execution to Codex **live, inside the current Claude session**. Produces an `approved` packet via the shared contract (same helpers `\u002Fhandoff --target=codex` uses), then invokes Codex synchronously with `codex exec \"\u003Ctarget-skill> --execute-approved\"`. Claude stays the orchestrator; Codex executes; Claude resumes once Codex returns.\n\nUnlike `\u002Fhandoff --target=codex` (async — user resumes later), `\u002Fdelegate` is live cross-CLI transport. The target skill defaults to `$exec`; other executor skills (e.g. `$ship`) may be named but `$exec` is the primary use case.\n\n## Mode requirement\n\n`\u002Fdelegate` is **`hybrid`-only by design**. `claude-only` has no executor to delegate to; `codex-only` plans in Codex directly. In either non-hybrid mode, the skill exits non-zero with a `mode-mismatch:` reason and touches no packet state.\n\n## Transport dependency\n\n`\u002Fdelegate` requires the `codex` binary on `PATH`. If missing (or auth-failed, or the start marker never prints), the skill falls cleanly into the **pre-start-failure** branch of the fallback matrix — it does not crash, and the packet stays at `approved`.\n\n## Process\n\n1. **Resolve agent mode** via `.\u002Fscripts\u002Fagent-mode.sh`. If the effective mode is anything other than `hybrid`, exit non-zero with a `mode-mismatch:` message naming the resolved mode. Do not touch the packet.\n\n2. **Parse arguments** from `$ARGUMENTS`:\n   - First positional token (if any) is the target skill; defaults to `$exec`.\n   - Collect repeatable `--allow-dirty \u003Cglob>` flags (shell-glob semantics, same as `scripts\u002Fapproved-plan.sh check`). Note: back-to-back hybrid cycles rewrite the `tasks\u002Fapproved-plan.md` mirror on each `consume`; commit it between cycles, or pass `--allow-dirty tasks\u002Fapproved-plan.md` so the next `draft` tolerates the uncommitted mirror.\n   - `--inline-fallback` auto-selects inline Claude execution in the pre-start-failure branch without prompting.\n\n3. **Derive `phase` \u002F `step` \u002F `title` from `tasks\u002Ftodo.md`**, using the same rules as `\u002Fhandoff --target=codex`:\n   - First unchecked `- [ ]` under the `### Active Step Plan` block.\n   - Fallback: first unchecked `- [ ]` under the current `## Phase N` header.\n   - Parse `Phase N` and `Step N.X` tokens from the surrounding context; the checkbox text (stripped of `- [ ] **Step N.X** — `) becomes `title`.\n\n4. **Produce the approval packet**:\n   1. `.\u002Fscripts\u002Fapproved-plan.sh draft --phase \"Phase N\" --step \"Step N.X\" --title \"\u003Ctitle>\"` plus any `--allow-dirty \u003Cglob>` flags. Surface the helper's single-line failure reason verbatim if it fails.\n   2. Pretty-print the drafted packet (`jq . .agents\u002Fapproved-plan.json`).\n   3. Ask exactly one concise question: *\"Approve this packet for Codex execution via \u002Fdelegate?\"*\n   4. On yes → `.\u002Fscripts\u002Fapproved-plan.sh approve`. On no → leave at `draft` and stop; the user can approve later or discard with `.\u002Fscripts\u002Fapproved-plan.sh supersede`.\n\n5. **Invoke Codex synchronously**:\n   - Record a start-marker timestamp (ISO-8601 UTC) before the exec call so the fallback matrix can classify outcomes.\n   - Shell out to `codex exec \"\u003Ctarget-skill> --execute-approved\"`. Capture stdout, stderr, and the exit code.\n   - The Step 4 consumer (`$exec --execute-approved`) is what actually runs `scripts\u002Fapproved-plan.sh check && consume` inside Codex. `\u002Fdelegate` is transport only.\n\n6. **Classify via the safe-fallback matrix — never blind-retry**:\n\n   - **Pre-start failure** — `codex` binary not found, auth failure, or the start marker was never printed by Codex before exit:\n     - Packet stays at `approved` (no writer ran).\n     - Offer two options: (a) run inline in Claude (same session, no CLI boundary), or (b) keep the packet `approved` for later manual `$exec --execute-approved` in Codex.\n     - If `--inline-fallback` was passed, auto-select (a) without prompting.\n     - Under an `agent-team` execution profile, inline fallback is a **downgrade** requiring explicit interactive confirmation unless `--inline-fallback` is already on the command line.\n\n   - **Success** — Codex exited 0, printed `Approved packet consumed: …`, and `tasks\u002Fapproved-plan.md` reports `lifecycle: consumed`:\n     - Step 4's `consume` already flipped `approved → consumed`.\n     - Read `tasks\u002Fapproved-plan.md` to confirm `lifecycle: consumed`. If confirmed, report success and hand back to the user.\n     - If the mirror does not report `consumed`, treat as the ambiguous case below.\n\n   - **Ambiguous** — non-zero exit, timeout, or crash after the start marker printed (Codex may have started mutating state):\n     - Call `.\u002Fscripts\u002Fapproved-plan.sh mark-uncertain`. The packet lifecycle becomes `uncertain`.\n     - Prompt the user with three explicit options: *inspect repo state manually \u002F discard the packet via `.\u002Fscripts\u002Fapproved-plan.sh supersede` \u002F continue inline in Claude from the current repo state*.\n     - **Never blind-retry `codex exec` on the same packet.** Re-executing against a packet whose consumer may have partially mutated state risks double-execution.\n\n7. **Report** the classified outcome, the final lifecycle, and whatever the user chose from the fallback prompt.\n\n## Constraints\n\n- **Hybrid-only.** Non-hybrid modes exit with `mode-mismatch:` before any packet operation.\n- **Never blind-retry cross-CLI.** The invariant. Ambiguous failures always go through `mark-uncertain` + explicit user prompt.\n- **Contract is frozen.** Do not edit `docs\u002Foperating-modes.md`, `specs\u002Fapproved-plan.schema.json`, or add new lifecycle states \u002F freshness checks. `uncertain` is already in the schema enum; Step 6 is just its first writer.\n- **`mark-uncertain` is only valid from `approved`.** The helper rejects `draft`, `consumed`, `stale`, `superseded`, and `uncertain` source states.\n- **`codex` binary is a transport dependency.** Fall cleanly into pre-start-failure if missing; do not crash.\n\n## Alignment Page\n\nFollow the shared alignment-page convention via the packaged convention resolver; output path is `alignment\u002Fdelegate-{topic}.html`.\n\n## Default Shipping Contract\n\nFollow the shared shipping contract convention in CLAUDE.md.\n",{"name":5,"type":25,"version":40,"description":7,"argument-hint":41,"required_conventions":42},"v0.1","[target-skill] [--allow-dirty \u003Cglob>] [--inline-fallback]",[43],"alignment-page","master","96e2912af886dd1606b1b97df707715aef062b02","packs\u002Fagent-bridge\u002Fclaude\u002Fdelegate",0,"https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills","https:\u002F\u002Fagentic-skills-blush.vercel.app","https:\u002F\u002Fgithub.com\u002FGeorgeQLe","georgeqle.com","George Le is a futurist and serial tech entrepreneur who wants to work toward a world where humanity is free from robotic tasks.","2026-11-10 04:07:24",{"source":8,"skillName":5,"commands":55,"prompt":65,"zipAvailable":66},[56,59,62],{"tool":57,"command":58},"npx","npx skills add GeorgeQLe\u002Fagentic-skills --skill delegate",{"tool":60,"command":61},"bunx","bunx skills add GeorgeQLe\u002Fagentic-skills --skill delegate",{"tool":63,"command":64},"pnpm","pnpm dlx skills add GeorgeQLe\u002Fagentic-skills --skill delegate","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Fpacks\u002Fagent-bridge\u002Fclaude\u002Fdelegate\u002FSKILL.md",true,[68,74,78],{"path":69,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":72,"sourceHtmlUrl":73},"ALIGNMENT-PAGE.md","REFERENCE","参考文档",52396,"https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Fpacks\u002Fagent-bridge\u002Fclaude\u002Fdelegate\u002FALIGNMENT-PAGE.md",{"path":75,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":76,"sourceHtmlUrl":77},"CHANGELOG.md",311,"https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Fpacks\u002Fagent-bridge\u002Fclaude\u002Fdelegate\u002FCHANGELOG.md",{"path":79,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":80,"sourceHtmlUrl":81},"references\u002Flanguages.md",4786,"https:\u002F\u002Fgithub.com\u002Ffadrienne\u002FMyGitHub\u002Fblob\u002Fmain\u002Fopenscience\u002Fbackend\u002Fcli\u002Fskills\u002Fllm-tools\u002Fwhisper\u002Freferences\u002Flanguages.md",[83,95,105,116,130,141],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":35},"01M6G0NJES15RJKZE6YZ0TSYGF","source-checkout-allowed","Valid source-checkout-only fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fsource-checkout-allowed\u002FSKILL.md",[],[90,91,92,93,28,94],"source checkout allowed","checkout","fixture","only","valid",{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":35},"01M6G0NFZ3R1MNB1E3DVTM0WB7","deck-install-valid","Valid deck install fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fdeck-install\u002FSKILL.md",[],[102,103,104,94,92,28],"deck install valid","deck","install",{"skillNo":106,"name":107,"nameZh":6,"description":108,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":109,"categories":110,"keywords":111,"publishedAt":35},"01M6G0ND3QZQC4MDZ9NVB97PT2","codex-pack-guard-valid","Valid Codex Pack Availability Guard fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fcodex-pack-guard\u002FSKILL.md",[],[112,113,114,94,24,115,92,28],"codex pack guard valid","pack","guard","availability",{"skillNo":117,"name":118,"nameZh":6,"description":119,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":120,"categories":121,"keywords":127,"publishedAt":35},"01M6G0NA9YB3NKGZKD3S96JMNB","codex-missing-skill-valid","Valid Codex missing skill fallback fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fcodex-missing-skill-fallback\u002FSKILL.md",[122],{"categoryNo":123,"categoryName":124,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[128,129,94,24,28,32,92],"codex missing skill valid","missing",{"skillNo":131,"name":132,"nameZh":6,"description":133,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":134,"categories":135,"keywords":137,"publishedAt":35},"01M6G0N7PC0EKFD40ZWQAXSAQ9","codex-base-skill-fallback-valid","Valid Codex base skill fallback fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fcodex-base-skill-fallback\u002FSKILL.md",[136],{"categoryNo":123,"categoryName":124,"taxonomyType":125,"taxonomyTypeDesc":126},[94,24,32,138,139,140,28,92],"base skill","codex base skill fallback valid","base",{"skillNo":142,"name":143,"nameZh":6,"description":144,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":145,"categories":146,"keywords":147,"publishedAt":35},"01M6G0N4VRR2YKZQ6N3Y6S2BQA","claude-pack-guard-valid","Valid Claude Pack Availability Guard fixture.","https:\u002F\u002Fgithub.com\u002FGeorgeQLe\u002Fagentic-skills\u002Fblob\u002Fmaster\u002Ftests\u002Ffixtures\u002Fskill-install-routing\u002Fvalid\u002Fclaude-pack-guard\u002FSKILL.md",[],[94,114,113,148,115,28,92],"claude pack guard valid"]