{"id":"740ea101-d9de-482e-8b39-d40ecefc9e2c","shortId":"3pMPbV","kind":"skill","title":"ask-codex","tagline":">-","description":"# Claude Code × Codex Collaboration Framework\n\nClaude Code is the architect and coordinator; Codex is the autonomous implementer.\n\n**Core principle:** Claude Code handles user interaction, scope control, product/architecture decisions, and final verification. Codex handles heavyweight codebase exploration, data inspection, implementation, and command execution. Default to Codex for repo-bound heavy lifting; keep Claude Code focused on judgment and synthesis. For anything that benefits from ongoing observation or mid-course correction, use the brokered session flow instead of one-shot execution.\n\n## Iron Laws\n\n1. **Route first** — On receiving an implementation task, determine routing (self vs delegate vs split) before acting.\n2. **Review loop** — Every Codex output must be verified by Claude Code (at minimum: read changed files). Never blindly trust results.\n3. **Self-contained context** — Prompts sent to Codex must be fully self-contained. Never rely on \"it should know.\"\n4. **Prefer session reuse** — When corrections are needed, prefer `--session` to continue in the original session rather than starting fresh and losing context.\n5. **Never hide failures** — When Codex fails, report honestly to the user with failure analysis. No silent retries.\n6. **Auto-offload heavy lifting** — If a subtask is mostly repo/data exploration or verification work, prefer Codex by default instead of Claude Code or an internal lightweight/Haiku-style handoff.\n\n## Critical Rules\n\n- Use `~/.claude/skills/ask-codex/scripts/ask_codex.sh` for single-turn execution.\n- Use `~/.claude/skills/ask-codex/scripts/codex_broker.sh` for live collaboration, long tasks, review loops, or any task where Claude Code may want to inspect progress and send follow-up instructions before the whole effort is over.\n- Do not call the `codex` CLI directly from the skill workflow (except `codex exec review` — see `references/invocation.md`).\n- For one-shot `ask_codex.sh` runs: if it succeeds (exit code 0), read the output file. Don't re-run just because output seems short — Codex often works quietly.\n- Quote file paths containing `[`, `]`, spaces, or special characters.\n- **Keep prompts focused on goals and constraints, not implementation steps.** Aim for ~500 words max.\n- **Never paste file contents into the prompt.** Use `--file` to point Codex to files.\n- **Never mention this skill or its configuration in the prompt.**\n- **Require evidence, not assertions.** End every prompt with a `Verification:` line that names a concrete command Codex must run and show output for (tests, reproduction, grep, etc.). Do not accept \"tests pass\" — demand the run. This is the single biggest driver of Codex success rate.\n\nPrompt-design principles, templates, and anti-patterns live in `references/prompt-engineering.md`.\n\n---\n\n## Phase 1: Smart Task Routing\n\nAfter receiving a task, first determine the execution route. This is the entry logic for this framework.\n\n### Routing Decision Tree\n\n```\nTask received\n  │\n  ├─ User explicitly requested Codex?\n  │    → Respect user intent, delegate directly (skip to Phase 2)\n  │\n  ├─ Heavy internal subtask? (repo sweep, data scan, log triage, broad search, impact analysis,\n  │  evidence gathering, long verification run)\n  │    → Delegate to Codex automatically (usually Mode 0 / read-only)\n  │\n  ├─ Analysis / planning / decision task?\n  │    ├─ Mostly user-facing judgment? (architecture choice, trade-off discussion, requirements)\n  │    │    → Claude Code handles it. Do not delegate.\n  │    └─ Mostly evidence gathering from repo/data?\n  │         → Delegate to Codex, then Claude Code synthesizes\n  │\n  ├─ Git / PR / deployment operation? (commit, push, PR, release)\n  │    → Claude Code handles it. Codex lacks these tools.\n  │\n  ├─ Clear implementation task?\n  │    ├─ Requirements clear + self-contained description + pure code/shell?\n  │    │    ├─ User-visible execution choice matters?\n  │    │    │    → Suggest delegating to Codex\n  │    │    └─ Internal execution detail only?\n  │    │         → Delegate to Codex automatically\n  │    ├─ Requirements vague or need multi-turn user clarification?\n  │    │    → Claude Code clarifies first, consider delegation after\n  │    └─ Depends on current conversation context / transient state?\n  │         → Claude Code handles it (context can't transfer to Codex)\n  │\n  └─ Review / verification task?\n       → Choose cross-review mode based on source\n```\n\n### Routing Criteria Quick Reference\n\n| Criterion | → Claude Code | → Codex |\n|-----------|:---:|:---:|\n| Requires user interaction / confirmation | ✓ | |\n| Uncertain approach, multiple possibilities | ✓ | |\n| Involves git / PR / deployment | ✓ | |\n| Depends on current conversation context | ✓ | |\n| Heavy repo exploration across many files | | ✓ |\n| Data/log inspection and summarization | | ✓ |\n| Broad symbol/callsite/impact tracing | | ✓ |\n| Multi-command verification / reproduction | | ✓ |\n| Looks like an internal lightweight/Haiku subtask | | ✓ |\n| Clear requirements, well-defined goal | | ✓ |\n| Pure code read/write + shell operations | | ✓ |\n| Can be self-contained in <500 words | | ✓ |\n| Localized file/module changes | | ✓ |\n| Multiple independent subtasks | | ✓ (parallel) |\n\n### Automatic Codex-First Exception\n\nClaude Code may delegate **without asking the user first** when all of the following are true:\n\n- The delegated work is an internal execution detail, not a user-visible scope change\n- The subtask is primarily repo/data/tooling work rather than product judgment\n- Claude Code can package the task self-contained\n- Codex results will still be reviewed by Claude Code before reporting back\n\nThis exception is the default path for:\n\n- Codebase exploration spanning multiple modules, directories, or 5+ likely files\n- Searching call sites, dependencies, ownership, or impact radius\n- Inspecting logs, CSV/JSON/SQLite data, generated artifacts, or benchmark outputs\n- Running reproduction/verification commands and collecting evidence\n- Any background task that Claude Code would otherwise be tempted to push to a lightweight internal model such as Haiku\n\n### User-Visible Delegation vs Internal Delegation\n\nWhen delegation changes how the task is executed in a way the user likely cares about, surface it. When delegation is just internal heavy lifting, do it automatically.\n\nCorrect approach for user-visible delegation:\n\n```\n✓ \"This task is a good fit for Codex — [reason]. Shall I delegate?\"\n✓ \"I'll handle the analysis and planning; Codex can implement. Sound good?\"\n✗ (Silently route a major user-facing decision or deliverable change)\n```\n\nCorrect approach for internal heavy lifting:\n\n```\n✓ Claude Code silently uses Codex to scan the repo, inspect data, or run lengthy verification\n✓ Claude Code receives evidence from Codex, reviews it, then answers the user directly\n✗ Claude Code burns time doing broad exploration itself or routes it to a weaker lightweight model first\n```\n\nException: User has explicitly said \"use codex\", \"let codex do it\", etc. — delegate directly, no confirmation needed.\n\n---\n\n## Phase 2: Collaboration Mode Selection\n\nChoose a mode based on task characteristics. The one-line summary is here; full diagrams, examples, and constraints are in `references/collaboration-modes.md`.\n\n| Mode | When to use | Key script |\n|------|-------------|-----------|\n| **0 — Scout / Exploration** | Read-heavy fact-finding (search, logs, impact mapping). Default for any broad repo/data sweep Claude Code was about to do itself. | `ask_codex.sh --read-only` |\n| **A — Single Delegation** | Clearly bounded standalone implementation. | `ask_codex.sh` |\n| **B — Parallel Delegation** | Multiple independent subtasks with no file overlap (max 3 in parallel). | `ask_codex.sh` × N background |\n| **C — Iterative Review Loop** | High-quality-bar work; Claude Code reviews and sends specific fix instructions over `--session` (max 3 rounds). | `ask_codex.sh` + `--session` |\n| **D — Feasibility Check** | Validate a Claude-Code-designed approach against real code before committing to it. | `ask_codex.sh --read-only` |\n| **E — Cross Review** | Independent reviewer perspective (Codex reviews Claude's work, or Claude reviews Codex's). | `ask_codex.sh --read-only` or `codex exec review` |\n| **F — Brokered Collaboration** | Long-running work where Claude Code wants to inspect progress and steer mid-flight. | `codex_broker.sh` |\n\n**Heuristic:** If the task is short and self-contained → A. If it needs live observation or mid-course correction → F. If it's pure evidence-gathering → 0. If quality bar is high → C.\n\n---\n\n## Phase 3: Core Invocation Flow\n\nThis is the minimum you need to call Codex. Full option reference, `codex exec review`, and integration-with-other-skills table are in `references/invocation.md`.\n\n### Single-turn call\n\n```bash\n~/.claude/skills/ask-codex/scripts/ask_codex.sh \"Goal description\" \\\n  --file <entry-file-1> \\\n  --file <entry-file-2>\n```\n\n**Output on success:**\n\n```\nsession_id=<thread_id>\noutput_path=<path to markdown file>\n```\n\nRead `output_path` for Codex's response. Save `session_id` for follow-ups via `--session <sid>`.\n\n**Common role flags:**\n\n- `--read-only` → scout, reviewer, feasibility check\n- `--reasoning high` → debugger, hard refactors\n- `--session <sid>` → continue a previous session (same role only; see `references/session-management.md`)\n- `--timeout <s>` / `--idle-timeout <s>` → override defaults (600s / 180s)\n\n### Brokered call\n\n```bash\nbash ~/.claude/skills/ask-codex/scripts/codex_broker.sh start \"Goal description\" \\\n  --file <entry-file>\n# then: status / send / wait / stop against the returned broker_id\n```\n\n### Exit code handling (summary)\n\n| Exit | Action |\n|------|--------|\n| 0 | Read `output_path`, continue normal flow |\n| 4 | Codex is asking a question — answer via `--session <sid>` (max 5 relay rounds) |\n| 3 | Fatal error (connection/auth/service) — report to user, offer takeover; do NOT auto-retry |\n| 2 | Timeout after one automatic grace window — report partial output + offer split/retry/takeover |\n| 1 / 137 | Codex error / OOM — simplify prompt or split task |\n\nFull recovery decision tree, fast-fatal-error detection, review-loop failure handling, and takeover principles are in `references/failure-recovery.md`.\n\n---\n\n## Red Flags — STOP\n\nIf you catch yourself thinking any of these, stop immediately:\n\n- \"Codex should understand what I mean\" — No, the prompt must be self-contained and explicit.\n- \"The result looks roughly right\" — No, you must read the changed files to verify.\n- \"Let's skip review this time, we're in a hurry\" — No, the review loop is an Iron Law.\n- \"Let me just retry, it should work\" — No, analyze the failure reason before choosing a strategy.\n- \"This is only exploration, I can just do it myself first\" — No, heavy repo/data exploration should default to Codex.\n- \"This looks like a Haiku/background task\" — Good. Route it to Codex unless it requires user-facing judgment.\n- \"Just delegate, no need to ask the user\" — Only for internal heavy lifting or explicit user Codex requests.\n- \"Codex timed out earlier, so I should avoid delegating new work\" — No, independent new tasks can start a fresh session.\n- \"Let me include all the details in the prompt\" — No, 500 words max: goal + constraints + entry files.\n- \"One more review round should fix it\" — 3-round cap. Beyond that, stop and analyze.\n- \"Codex can't do it, I'll take over\" — Fine, but read Codex's partial output first.\n\n---\n\n## References\n\nLoad these on demand when the task needs the detail:\n\n- `references/collaboration-modes.md` — full diagrams, examples, constraints for Modes 0/A/B/C/D/E/F\n- `references/prompt-engineering.md` — principles, templates by task type, anti-patterns, context transfer, verification gates\n- `references/session-management.md` — role definitions, session lifecycle, timeout & new-task policy, resume limitations\n- `references/failure-recovery.md` — exit codes, multi-turn relay protocol, recovery decision tree, takeover principles\n- `references/invocation.md` — full options, `codex exec review`, integration with other skills","tags":["ask","codex","booksiyi1412","agent-skills","ai-agents","claude-code","codex-cli","skills","skills-sh"],"capabilities":["skill","source-booksiyi1412","skill-ask-codex","topic-agent-skills","topic-ai-agents","topic-claude-code","topic-codex","topic-codex-cli","topic-skills","topic-skills-sh"],"categories":["ask-codex"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/bookSiYi1412/ask-codex","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add bookSiYi1412/ask-codex","source_repo":"https://github.com/bookSiYi1412/ask-codex","install_from":"skills.sh"}},"qualityScore":"0.454","qualityRationale":"deterministic score 0.45 from registry signals: · indexed on github topic:agent-skills · 9 github stars · SKILL.md body (12,106 chars)","verified":false,"liveness":"unknown","lastLivenessCheck":null,"agentReviews":{"count":0,"score_avg":null,"cost_usd_avg":null,"success_rate":null,"latency_p50_ms":null,"narrative_summary":null,"summary_updated_at":null},"enrichmentModel":"deterministic:skill-github:v1","enrichmentVersion":1,"enrichedAt":"2026-05-18T19:08:34.446Z","embedding":null,"createdAt":"2026-05-09T07:03:32.184Z","updatedAt":"2026-05-18T19:08:34.446Z","lastSeenAt":"2026-05-18T19:08:34.446Z","tsv":"'/.claude/skills/ask-codex/scripts/ask_codex.sh':220,1196 '/.claude/skills/ask-codex/scripts/codex_broker.sh':227,1261 '0':286,473,981,1154,1282 '0/a/b/c/d/e/f':1578 '1':88,410,1328 '137':1329 '180s':1256 '2':105,448,949,1316 '3':126,1030,1056,1162,1302,1535 '4':147,1289 '5':170,757,1299 '500':325,667,1521 '6':188 '600s':1255 'accept':381 'across':629 'act':104 'action':1281 'aim':323 'analysi':184,461,477,861 'analyz':1428,1542 'answer':910,1295 'anti':404,1586 'anti-pattern':403,1585 'anyth':64 'approach':614,839,881,1069 'architect':13 'architectur':486 'artifact':773 'ask':2,686,1292,1478 'ask-codex':1 'ask_codex.sh':279,1007,1018,1033,1058,1077,1097 'assert':355 'auto':190,1314 'auto-offload':189 'auto-retri':1313 'automat':470,556,676,837,1320 'autonom':19 'avoid':1498 'b':1019 'back':742 'background':784,1035 'bar':1043,1157 'base':598,956 'bash':1195,1259,1260 'benchmark':775 'benefit':66 'beyond':1538 'biggest':391 'blind':123 'bound':52,1015 'broad':458,636,919,997 'broker':77,1106,1257,1274 'burn':916 'c':1036,1160 'call':260,761,1173,1194,1258 'cap':1537 'care':824 'catch':1363 'chang':120,671,711,812,879,1397 'charact':312 'characterist':959 'check':1062,1233 'choic':487,543 'choos':593,953,1433 'clarif':565 'clarifi':568 'claud':4,9,23,56,115,210,239,493,509,520,566,580,606,681,722,738,787,886,901,914,1000,1045,1066,1089,1093,1113 'claude-code-design':1065 'clear':528,532,650,1014 'cli':263 'code':5,10,24,57,116,211,240,285,494,510,521,567,581,607,657,682,723,739,788,887,902,915,1001,1046,1067,1072,1114,1277,1606 'code/shell':538 'codebas':38,750 'codex':3,6,16,35,48,109,134,175,205,262,270,301,339,368,394,439,469,507,524,548,555,589,608,678,731,852,864,890,906,937,939,1087,1095,1102,1174,1178,1212,1290,1330,1371,1454,1465,1489,1491,1543,1555,1620 'codex-first':677 'codex_broker.sh':1124 'collabor':7,230,950,1107 'collect':781 'command':44,367,641,779 'commit':516,1074 'common':1224 'concret':366 'configur':348 'confirm':612,946 'connection/auth/service':1305 'consid':570 'constraint':319,971,1525,1575 'contain':129,140,308,535,665,730,1134,1384 'content':331 'context':130,169,577,584,625,1588 'continu':158,1240,1286 'control':29 'convers':576,624 'coordin':15 'core':21,1163 'correct':74,152,838,880,1145 'cours':73,1144 'criteria':602 'criterion':605 'critic':217 'cross':595,1082 'cross-review':594 'csv/json/sqlite':770 'current':575,623 'd':1060 'data':40,454,771,896 'data/log':632 'debugg':1236 'decis':31,432,479,876,1340,1613 'default':46,207,747,994,1254,1452 'defin':654 'definit':1594 'deleg':100,443,467,499,505,546,553,571,684,698,806,809,811,829,844,856,943,1013,1021,1474,1499 'deliver':878 'demand':384,1564 'depend':573,621,763 'deploy':514,620 'descript':536,1198,1264 'design':399,1068 'detail':551,704,1516,1570 'detect':1346 'determin':96,419 'diagram':968,1573 'direct':264,444,913,944 'directori':755 'discuss':491 'driver':392 'e':1081 'earlier':1494 'effort':255 'end':356 'entri':426,1526 'error':1304,1331,1345 'etc':378,942 'everi':108,357 'evid':353,462,501,782,904,1152 'evidence-gath':1151 'exampl':969,1574 'except':269,680,744,931 'exec':271,1103,1179,1621 'execut':45,85,225,421,542,550,703,817 'exit':284,1276,1280,1605 'explicit':437,934,1386,1487 'explor':39,200,628,751,920,983,1439,1450 'f':1105,1146 'face':484,875,1471 'fact':988 'fact-find':987 'fail':176 'failur':173,183,1350,1430 'fast':1343 'fast-fatal-error':1342 'fatal':1303,1344 'feasibl':1061,1232 'file':121,290,306,330,336,341,631,759,1027,1199,1200,1265,1398,1527 'file/module':670 'final':33 'find':989 'fine':1552 'first':90,418,569,679,689,930,1446,1559 'fit':850 'fix':1051,1533 'flag':1226,1359 'flight':1123 'flow':79,1165,1288 'focus':58,315 'follow':249,694,1220 'follow-up':248,1219 'framework':8,430 'fresh':166,1509 'full':967,1175,1338,1572,1618 'fulli':137 'gate':1591 'gather':463,502,1153 'generat':772 'git':512,618 'goal':317,655,1197,1263,1524 'good':849,868,1461 'grace':1321 'grep':377 'haiku':802 'haiku/background':1459 'handl':25,36,495,522,582,859,1278,1351 'handoff':216 'hard':1237 'heavi':53,192,449,626,833,884,986,1448,1484 'heavyweight':37 'heurist':1125 'hide':172 'high':1041,1159,1235 'high-quality-bar':1040 'honest':178 'hurri':1411 'id':1205,1217,1275 'idl':1251 'idle-timeout':1250 'immedi':1370 'impact':460,766,992 'implement':20,42,94,321,529,866,1017 'includ':1513 'independ':673,1023,1084,1503 'inspect':41,244,633,768,895,1117 'instead':80,208 'instruct':251,1052 'integr':1183,1623 'integration-with-other-skil':1182 'intent':442 'interact':27,611 'intern':214,450,549,647,702,798,808,832,883,1483 'invoc':1164 'involv':617 'iron':86,1418 'iter':1037 'judgment':60,485,721,1472 'keep':55,313 'key':979 'know':146 'lack':525 'law':87,1419 'lengthi':899 'let':938,1401,1420,1511 'lifecycl':1596 'lift':54,193,834,885,1485 'lightweight':797,928 'lightweight/haiku':648 'lightweight/haiku-style':215 'like':645,758,823,1457 'limit':1603 'line':362,963 'live':229,406,1139 'll':858,1549 'load':1561 'local':669 'log':456,769,991 'logic':427 'long':231,464,1109 'long-run':1108 'look':644,1389,1456 'loop':107,234,1039,1349,1415 'lose':168 'major':872 'mani':630 'map':993 'matter':544 'max':327,1029,1055,1298,1523 'may':241,683 'mean':1376 'mention':343 'mid':72,1122,1143 'mid-cours':71,1142 'mid-flight':1121 'minimum':118,1169 'mode':472,597,951,955,975,1577 'model':799,929 'modul':754 'most':198,481,500 'multi':562,640,1608 'multi-command':639 'multi-turn':561,1607 'multipl':615,672,753,1022 'must':111,135,369,1380,1394 'n':1034 'name':364 'need':154,560,947,1138,1171,1476,1568 'never':122,141,171,328,342 'new':1500,1504,1599 'new-task':1598 'normal':1287 'observ':69,1140 'offer':1309,1326 'offload':191 'often':302 'one':83,277,962,1319,1528 'one-lin':961 'one-shot':82,276 'ongo':68 'oom':1332 'oper':515,660 'option':1176,1619 'origin':161 'otherwis':790 'output':110,289,298,373,776,1201,1206,1209,1284,1325,1558 'overlap':1028 'overrid':1253 'ownership':764 'packag':725 'parallel':675,1020,1032 'partial':1324,1557 'pass':383 'past':329 'path':307,748,1207,1210,1285 'pattern':405,1587 'perspect':1086 'phase':409,447,948,1161 'plan':478,863 'point':338 'polici':1601 'possibl':616 'pr':513,518,619 'prefer':148,155,204 'previous':1242 'primarili':715 'principl':22,400,1354,1580,1616 'product':720 'product/architecture':30 'progress':245,1118 'prompt':131,314,334,351,358,398,1334,1379,1519 'prompt-design':397 'protocol':1611 'pure':537,656,1150 'push':517,794 'qualiti':1042,1156 'question':1294 'quick':603 'quiet':304 'quot':305 'radius':767 'rate':396 'rather':163,718 're':294,1408 're-run':293 'read':119,287,475,985,1009,1079,1099,1208,1228,1283,1395,1554 'read-heavi':984 'read-on':474,1008,1078,1098,1227 'read/write':658 'real':1071 'reason':853,1234,1431 'receiv':92,415,435,903 'recoveri':1339,1612 'red':1358 'refactor':1238 'refer':604,1177,1560 'references/collaboration-modes.md':974,1571 'references/failure-recovery.md':1357,1604 'references/invocation.md':274,1190,1617 'references/prompt-engineering.md':408,1579 'references/session-management.md':1248,1592 'relay':1300,1610 'releas':519 'reli':142 'repo':51,452,627,894 'repo-bound':50 'repo/data':199,504,998,1449 'repo/data/tooling':716 'report':177,741,1306,1323 'reproduct':376,643 'reproduction/verification':778 'request':438,1490 'requir':352,492,531,557,609,651,1468 'respect':440 'respons':1214 'result':125,732,1388 'resum':1602 'retri':187,1315,1423 'return':1273 'reus':150 'review':106,233,272,590,596,736,907,1038,1047,1083,1085,1088,1094,1104,1180,1231,1348,1404,1414,1530,1622 'review-loop':1347 'right':1391 'role':1225,1245,1593 'rough':1390 'round':1057,1301,1531,1536 'rout':89,97,413,422,431,601,870,923,1462 'rule':218 'run':280,295,370,386,466,777,898,1110 'said':935 'save':1215 'scan':455,892 'scope':28,710 'scout':982,1230 'script':980 'search':459,760,990 'see':273,1247 'seem':299 'select':952 'self':98,128,139,534,664,729,1133,1383 'self-contain':127,138,533,663,728,1132,1382 'send':247,1049,1268 'sent':132 'session':78,149,156,162,1054,1059,1204,1216,1223,1239,1243,1297,1510,1595 'shall':854 'shell':659 'short':300,1130 'shot':84,278 'show':372 'silent':186,869,888 'simplifi':1333 'singl':223,390,1012,1192 'single-turn':222,1191 'site':762 'skill':267,345,1186,1626 'skill-ask-codex' 'skip':445,1403 'smart':411 'sound':867 'sourc':600 'source-booksiyi1412' 'space':309 'span':752 'special':311 'specif':1050 'split':102,1336 'split/retry/takeover':1327 'standalon':1016 'start':165,1262,1507 'state':579 'status':1267 'steer':1120 'step':322 'still':734 'stop':1270,1360,1369,1540 'strategi':1435 'subtask':196,451,649,674,713,1024 'succeed':283 'success':395,1203 'suggest':545 'summar':635 'summari':964,1279 'surfac':826 'sweep':453,999 'symbol/callsite/impact':637 'synthes':511 'synthesi':62 'tabl':1187 'take':1550 'takeov':1310,1353,1615 'task':95,232,237,412,417,434,480,530,592,727,785,815,846,958,1128,1337,1460,1505,1567,1583,1600 'templat':401,1581 'tempt':792 'test':375,382 'think':1365 'time':917,1406,1492 'timeout':1249,1252,1317,1597 'tool':527 'topic-agent-skills' 'topic-ai-agents' 'topic-claude-code' 'topic-codex' 'topic-codex-cli' 'topic-skills' 'topic-skills-sh' 'trace':638 'trade':489 'trade-off':488 'transfer':587,1589 'transient':578 'tree':433,1341,1614 'triag':457 'true':696 'trust':124 'turn':224,563,1193,1609 'type':1584 'uncertain':613 'understand':1373 'unless':1466 'up':1221 'use':75,219,226,335,889,936,978 'user':26,181,436,441,483,540,564,610,688,708,804,822,842,874,912,932,1308,1470,1480,1488 'user-fac':482,873,1469 'user-vis':539,707,803,841 'usual':471 'vagu':558 'valid':1063 'verif':34,202,361,465,591,642,900,1590 'verifi':113,1400 'via':1222,1296 'visibl':541,709,805,843 'vs':99,101,807 'wait':1269 'want':242,1115 'way':820 'weaker':927 'well':653 'well-defin':652 'whole':254 'window':1322 'without':685 'word':326,668,1522 'work':203,303,699,717,1044,1091,1111,1426,1501 'workflow':268 'would':789","prices":[{"id":"95a2c729-c758-4027-a81d-6a6a89905e6f","listingId":"740ea101-d9de-482e-8b39-d40ecefc9e2c","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"bookSiYi1412","category":"ask-codex","install_from":"skills.sh"},"createdAt":"2026-05-09T07:03:32.184Z"}],"sources":[{"listingId":"740ea101-d9de-482e-8b39-d40ecefc9e2c","source":"github","sourceId":"bookSiYi1412/ask-codex","sourceUrl":"https://github.com/bookSiYi1412/ask-codex","isPrimary":false,"firstSeenAt":"2026-05-09T07:03:32.184Z","lastSeenAt":"2026-05-18T19:08:34.446Z"}],"details":{"listingId":"740ea101-d9de-482e-8b39-d40ecefc9e2c","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"bookSiYi1412","slug":"ask-codex","github":{"repo":"bookSiYi1412/ask-codex","stars":9,"topics":["agent-skills","ai-agents","claude-code","codex","codex-cli","skills","skills-sh"],"license":"apache-2.0","html_url":"https://github.com/bookSiYi1412/ask-codex","pushed_at":"2026-04-23T14:27:32Z","description":"A Claude Code skill that helps Claude Code delegate repository work to the Codex CLI","skill_md_sha":"1653f9356620f42c04533c805b800d93cfcbdc3f","skill_md_path":"SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/bookSiYi1412/ask-codex"},"layout":"root","source":"github","category":"ask-codex","frontmatter":{"name":"ask-codex","description":">-"},"skills_sh_url":"https://skills.sh/bookSiYi1412/ask-codex"},"updatedAt":"2026-05-18T19:08:34.446Z"}}