{"id":"a256a02a-5efe-41c8-bb5c-1789a0094f5d","shortId":"e3W6EX","kind":"skill","title":"autogrind","tagline":"Engage in 24x7 auto-work mode. Continuously grind through improvements, fixes, tests, and polish without stopping until you say so. For long-running sessions across code, ML/data, research, design, or writing. Not for single bounded tasks. Trigger phrases: /autogrind, /自己动, 'auto","description":"# AutoGrind\n\n## Overview\n\nAutoGrind keeps the agent continuously working through a five-phase cycle: Overview → Understand → Plan → Work → Reflect → 60s pause → repeat. The agent never decides the project is \"done enough.\" Only the user decides when to stop.\n\n**Not for single tasks or interactive work.** AutoGrind is a mode, not a command. Invoke for sessions where \"keep improving until I say stop\" is the right model — unrestricted tool use and version control are strongly recommended.\n\n## The Iron Law\n\n```\nGRIND UNTIL EXPLICIT STOP SIGNAL\n```\n\n**Violating the letter of this rule is violating the spirit of this rule.**\n\n- Completing all current tasks is **NOT** a stop condition\n- \"Everything looks good\" is **NOT** a stop condition\n- End of a cycle is **NOT** a stop condition\n\n## The Grind Cycle\n\n```dot\ndigraph autogrind {\n    rankdir=TB;\n    init      [label=\"INIT (once)\\nDetect guidance files\\nInit Session Heuristics\", shape=box];\n    overview  [label=\"1. OVERVIEW\\nAssess state · importance-rate areas\", shape=box];\n    understand[label=\"2. UNDERSTAND\\nReview relevant work & history\", shape=box];\n    plan      [label=\"3. PLAN\\nPrioritized tasks · frontier scan\\nsolvability gate\", shape=box];\n    work      [label=\"4. WORK\\nExecute · validate · persist\", shape=box];\n    reflect   [label=\"5. REFLECT\\nGrounded signals · pattern check\\nheuristic extraction\", shape=box];\n    pause     [label=\"PAUSE 60s\\nAnnounce · wait · continue\", shape=box, style=filled, fillcolor=\"#ffffcc\"];\n    check     [label=\"Explicit stop\\nsignal?\", shape=diamond];\n    done      [label=\"STOP\", shape=doublecircle];\n    warn      [label=\"NEVER stop\\non your own\", shape=box, style=filled, fillcolor=\"#ff4444\", fontcolor=white];\n\n    init -> overview;\n    overview -> understand -> plan -> work -> reflect -> pause -> check;\n    check -> done      [label=\"yes\"];\n    check -> overview  [label=\"no - always\"];\n    check -> warn      [label=\"tempted\\nto stop\"];\n}\n```\n\n## Discovery Mindset\n\nTreat every AutoGrind session as rigorous scientific inquiry — you are a brilliant scientist, fearless leader, and ruthless pioneer at the frontier.\n\n**Explore boldly.** Challenge every assumption. The most valuable finding is the one nobody was looking for.\n\n**Claim confidently.** Back every bold claim with state-of-the-art research and real measurements, not vague hedges.\n\n**Analyze from multiple angles.** Empirical, theoretical, structural, behavioral — each is a distinct lens. Synthesize; a single-angle finding is incomplete.\n\n**Explain at two levels.** Both intuitive (\"why does this make sense?\") and theoretical (\"what mechanism underlies it?\"). Insight without mechanism is folklore.\n\n**Pivot without hesitation.** Turbulence is signal, not failure. When evidence contradicts the strategy, update completely — replace bad designs; a drastic pivot beats a stubborn march toward a wrong answer.\n\n**Iterate toward understanding.** Each cycle is an experiment: hypothesize → implement → measure → conclude. Not done when things work — done when you understand *why*.\n\n## Workflow\n\n### INIT - once per session\n\n- Scan for guidance files: `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, `.cursorrules`, `opencode.md`, `README.md`\n- Extract: project goals, domain, methodology or tech stack, conventions, known issues\n- If none exist, infer from directory structure, existing artifacts, and project context\n- Initialize **Session Heuristics**: an empty in-context list (max 5) of transferable principles discovered during Reflect phases. Format: `[cycle N] When <condition>, prefer <approach> because <reason>.` Prepend each Overview with a quick read of this list.\n- **Context compaction**: complete the current phase and continue — each Overview re-reads state from scratch. Session Heuristics reinitialize to empty if lost.\n\n### Phase 1 - Overview\n\nAssess current project state. Adapt to domain:\n\n- **Code**: `git log --oneline -20`, `git status`, run test suite, scan `TODO`/`FIXME`\n- **ML/research**: review experiment log or training runs, check latest metrics, scan open questions\n- **Design/writing**: review revision history, open feedback, check revision backlog\n\nProduce a one-paragraph current-state summary. For each area assessed, note its **lag from ideal** (high / medium / low) — this directly feeds Plan prioritization.\n\nRead Session Heuristics before proceeding to Understand.\n\n### Phase 2 - Understand\n\n- Review artifacts most relevant to this cycle's focus (code, data, papers, designs, drafts)\n- Review recent changes; identify failing validations, open questions, broken areas\n- Do not start planning until understanding is solid\n\n### Phase 3 - Plan\n\n**Own the work.** What is the highest-leverage change right now? Reason from first principles — challenge assumptions, find non-obvious problems. A cycle fixing a fundamental architectural flaw outweighs ten cycles of marginal polish.\n\nGenerate 3–6 tasks. Fewer, well-scoped tasks beat long lists. Keep each task to **≤ 4 steps** for reliable execution. **Each task must produce a visible, verifiable output change.** Discard micro-tasks that could be grouped or that wouldn't stand alone as a commit — fold them into substantive ones. Priority order applies across all domains:\n\n1. Broken/failing validations — tests, failed experiments, broken builds\n2. Incomplete core deliverables — features, analyses, missing sections\n3. Quality/coverage gaps — test coverage, experiment coverage, argument gaps\n4. Documentation/writeup gaps\n5. Performance/efficiency opportunities\n6. Polish/refinement\n\n**Capability frontier**: after listing priority tasks, identify 1–2 frontier tasks — work that introduces something the project currently lacks: a capability not yet built, a property not yet measured, a path with no coverage. They will not appear on any existing TODO list.\n\n**Output bar**: at least one task must be discovered — a problem not on any TODO, a non-obvious improvement, or a deeper solution over an obvious patch. If all tasks were already listed, run the frontier scan at higher ambition.\n\n**Solvability gate**: verify each task is actionable. Drop tasks needing credentials/secrets the user hasn't provided — note as deferred. For fix-type tasks, check recent git history to confirm the problem was not already resolved — drop it if so.\n\nTrack tasks with the platform's native task mechanism.\n\n### Phase 4 - Work\n\n- Execute tasks in priority order\n- Execute **independent tasks concurrently** where supported\n- Per task: verify (confirm problem still exists — check git history, reproduce; if resolved, no change is the correct output) → execute → validate (tests, outputs, metrics) → persist (commit, checkpoint, log)\n- One logical change per persist — never batch unrelated changes\n- Git commits: use `git -c commit.gpgsign=false commit` (avoids signing prompts). Use semantic commit messages: `feat:`, `fix:`, `docs:`, `test:`, `chore:`, `refactor:`, `perf:`, `style:`\n- If blocked: note the blocker, skip to the next task\n- Interrupt the user only if **all** remaining tasks share the same unresolvable blocker\n- User message (any phase, any type): handle it immediately — answer, redirect, or incorporate — then announce `\"Resuming AutoGrind cycle [N]...\"` and continue the current phase. Steering ≠ cycle end.\n- Critical issue discovered mid-task (security flaw, data loss): add a FIXME with severity, continue planned tasks, and defer the fix to next cycle's Phase 3.\n- **Safety boundary**: stay within the project directory; do not modify system files, delete outside the project, or run operations that normally require human confirmation.\n- **Permission mode**: bypass permissions only — mode switches introduce approval prompts.\n\n### Phase 5 - Reflect\n\n**Step 1 — Grounded signals first.** Before any self-assessment, check verifiable evidence:\n\n- Code: test results, lint/build status, coverage delta\n- ML/research: metric movement vs. last cycle, experiment outcomes\n- Design/writing: reviewer feedback received, revision diff, checklist completion\n\nDo not skip to self-assessment — these facts anchor the reflection.\n\n**Step 2 — Answer the two mandatory questions first — they override all other priorities:**\n\n**Core deliverable check**: Did this cycle directly improve the PRIMARY OUTPUT (the skill, model, paper, design, feature)? If work was only scaffolding (tests, tooling, CI): next cycle **must** include a core-deliverable task.\n\n**Self-audit**: Am I fixing real problems or adapting to symptoms? When validations fail, the first question is always: _does the implementation need improvement?_ Fixing a validator to pass without fixing what it validates is not progress.\n\n**Step 3 — Scan remaining dimensions:**\n\n| Dimension                | Ask                                               |\n| ------------------------ | ------------------------------------------------- |\n| Validation coverage      | Are important scenarios and edge cases exercised? |\n| Error/edge-case handling | Are failure modes handled gracefully?             |\n| Documentation            | Complete, accurate, up to date?                   |\n| Performance              | Any obvious bottlenecks?                          |\n| UX / output              | Is feedback clear and helpful?                    |\n| Observability            | Is logging/reporting adequate?                    |\n| Security                 | Any obvious attack surfaces?                      |\n| Work quality             | Anything to simplify or clarify?                  |\n\n**Step 4 — Cross-cycle pattern check.** If the same dimension is flagged with the same diagnosis and no measurable progress — **stuck loop**. Next cycle: **Refresh** by leading with a different dimension from the Step 3 table; do not return until the refresh cycle closes a different gap.\n\n**Step 5 — Extract one heuristic:** `When <condition>, prefer <approach> because <reason>.` Prepend to Session Heuristics (max 5, drop oldest).\n\nEnd Reflect with: _\"Next cycle focus: [area].\"_\n\n### Inter-Cycle Pause\n\nAfter Reflect: print `\"Cycle [N] complete. Starting cycle [N+1] in 60 seconds — stop signal now to halt.\"`, wait 60s (`sleep 60`), then begin Overview. Not a stopping point. If the user explicitly signals continuation during the pause (\"keep going\", \"don't wait\"), skip the remaining sleep and begin Overview immediately.\n\n## Stopping Conditions\n\n**One and only one:** the user sends an explicit stop signal.\n\nRecognized (English): \"stop\", \"pause\", \"halt\", \"exit autogrind\", \"that's enough\", or a direct equivalent. Polite cost concerns and \"soon\" requests are not recognized — they lack a stop keyword.\nRecognized (中文): \"停\", \"停止\", \"暂停\", \"够了\", \"结束\", or any unambiguous 中文 termination request.\n**Stop mid-task:** finish the atomic task, print `\"AutoGrind stopped after cycle [N].\"`, then stop. **Stop during analysis phases** (Overview/Understand/Plan/Reflect) or inter-cycle pause: stop cleanly — these phases have no in-flight code changes. Follow-ups are regular tasks, not new sessions — only `/autogrind` restarts AutoGrind.\n\nEverything else — silence, task completion, praise, cost concerns, polite suggestions (\"I'd appreciate if you wrapped up soon\"), questions, inter-cycle pauses, \"looks done\" — is **not** a stop signal.\n\n## Red Flags — Continue Immediately\n\n- \"TODO list empty\" or \"no obvious next task\" → Capability frontier scan always finds one\n- \"Project looks complete\" or \"everything is working\" → Measure it: coverage, perf, docs\n- \"Good enough to ship\" or \"I've been at this a while\" → Only the user decides\n- \"I'll summarize progress and pause\" → Pausing IS stopping\n- \"User praised my work / seems happy\" → Satisfaction ≠ stop signal\n- \"User sent a message / I just addressed their request\" → Not done — announce `\"Resuming AutoGrind...\"` and continue immediately. Steering ≠ stop.\n- \"Tests/validations pass now\" → Passing confirms correctness; never a stop signal\n- \"I improved tests/tooling this cycle\" → Scaffolding ≠ core deliverable; next cycle targets the primary output\n- \"Critical bug found mid-work\" → Document with a FIXME+severity and continue; Phase 3 will prioritize the fix\n- \"Every task was already on a TODO/FIXME list\" → frontier scan at higher ambition; discover at least one task\n- \"Session ended — user gave me a follow-up task\" → Regular task. Only `/autogrind` restarts the session.\n\n## Common Rationalizations\n\n| Rationalization                             | Reality                                                                                                                   |\n| ------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- |\n| \"I should check in with the user\"           | Work. They'll stop you when they need to.                                                                                 |\n| \"The test/validator was wrong, I fixed it\"  | First ask: does the _implementation_ need improvement? Fixing evaluators to match broken implementations is not progress. |\n| \"I have a fix task, so I should patch it\"   | Verify the problem still exists — check git history, reproduce. Already resolved = no change is the correct output. |\n| \"I completed many tasks this cycle\"          | Count means nothing if outputs aren't visible and verifiable. Micro-tasks that wouldn't stand alone as commits are not progress. |\n| \"Context window filling up — should stop\"    | Each Overview re-reads project state. Compaction is handled; finish the phase.                                           |\n| \"Let me outline the plan before starting\"   | Procrastination. Phase 3 is the plan. Phase 4 executes immediately — no meta-planning step in between.                   |","tags":["autogrind","ttttonyhe","agent-skill","agent-skills","ai-agents","autonomous-agents","claude-code","cursor","gemini-cli","llm-agent","openai-codex","opencode"],"capabilities":["skill","source-ttttonyhe","skill-autogrind","topic-agent-skill","topic-agent-skills","topic-ai-agents","topic-autonomous-agents","topic-claude-code","topic-cursor","topic-gemini-cli","topic-llm-agent","topic-openai-codex","topic-opencode","topic-skills-sh"],"categories":["autogrind"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/ttttonyhe/autogrind","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add ttttonyhe/autogrind","source_repo":"https://github.com/ttttonyhe/autogrind","install_from":"skills.sh"}},"qualityScore":"0.453","qualityRationale":"deterministic score 0.45 from registry signals: · indexed on github topic:agent-skills · 7 github stars · SKILL.md body (13,749 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":"2026-05-18T19:13:32.793Z","embedding":null,"createdAt":"2026-05-18T13:20:54.101Z","updatedAt":"2026-05-18T19:13:32.793Z","lastSeenAt":"2026-05-18T19:13:32.793Z","tsv":"'+1':1391 '-20':567 '/autogrind':42,1534,1724 '1':189,554,763,803,1119 '2':201,632,771,804,1167 '24x7':4 '3':211,667,706,779,1080,1252,1342,1688,1855 '4':223,721,788,930,1308,1860 '5':232,506,791,1116,1356,1368 '6':707,794 '60':1393,1403 '60s':64,245,1401 'accur':1276 'across':28,760 'action':886 'adapt':560,1222 'add':1063 'address':1637 'adequ':1294 'agent':50,68 'agents.md':468 'alon':748,1821 'alreadi':871,914,1696,1790 'alway':299,1232,1582 'ambit':879,1705 'analys':776 'analysi':1505 'analyz':364 'anchor':1163 'angl':367,381 'announc':1040,1642 'answer':435,1035,1168 'anyth':1302 'appear':833 'appli':759 'appreci':1549 'approv':1113 'architectur':697 'area':196,609,657,1377 'aren':1809 'argument':786 'art':356 'artifact':492,635 'ask':1257,1756 'assess':556,610,1127,1160 'assumpt':333,686 'atom':1493 'attack':1298 'audit':1215 'auto':6,44 'auto-work':5 'autogrind':1,45,47,90,172,310,1042,1452,1496,1536,1644 'avoid':988 'back':347 'backlog':597 'bad':423 'bar':840 'batch':977 'beat':428,714 'begin':1405,1430 'behavior':371 'block':1004 'blocker':1007,1025 'bold':330,349 'bottleneck':1283 'bound':38 'boundari':1082 'box':186,198,208,220,229,241,250,275 'brilliant':319 'broken':656,769,1766 'broken/failing':764 'bug':1675 'build':770 'built':819 'bypass':1107 'c':984 'capabl':796,816,1579 'case':1265 'challeng':331,685 'chang':650,678,734,957,973,979,1523,1793 'check':237,255,290,291,295,300,583,595,904,950,1128,1181,1313,1734,1786 'checklist':1152 'checkpoint':969 'chore':999 'ci':1203 'claim':345,350 'clarifi':1306 'claude.md':467 'clean':1514 'clear':1288 'close':1351 'code':29,563,643,1131,1522 'command':96 'commit':751,968,981,987,993,1823 'commit.gpgsign':985 'common':1728 'compact':531,1840 'complet':141,421,532,1153,1275,1387,1541,1587,1799 'concern':1462,1544 'conclud':447 'concurr':940 'condit':149,157,166,1434 'confid':346 'confirm':909,946,1104,1654 'context':495,503,530,1827 'continu':9,51,248,537,1046,1068,1416,1569,1646,1686 'contradict':417 'control':116 'convent':481 'core':773,1179,1210,1666 'core-deliver':1209 'correct':960,1655,1796 'cost':1461,1543 'could':740 'count':1804 'coverag':783,785,829,1136,1259,1594 'credentials/secrets':890 'critic':1053,1674 'cross':1310 'cross-cycl':1309 'current':143,534,557,604,813,1048 'current-st':603 'cursorrul':470 'cycl':58,161,169,440,515,640,693,701,1043,1051,1077,1143,1184,1205,1311,1331,1350,1375,1380,1385,1389,1499,1511,1558,1664,1669,1803 'd':1548 'data':644,1061 'date':1279 'decid':70,79,1612 'deeper':861 'defer':898,1072 'delet':1093 'deliver':774,1180,1211,1667 'delta':1137 'design':32,424,646,1194 'design/writing':589,1146 'diagnosi':1323 'diamond':261 'diff':1151 'differ':1337,1353 'digraph':171 'dimens':1255,1256,1317,1338 'direct':620,1185,1458 'directori':489,1087 'discard':735 'discov':510,847,1055,1706 'discoveri':306 'distinct':375 'doc':997,1596 'document':1274,1680 'documentation/writeup':789 'domain':476,562,762 'done':74,262,292,449,453,1561,1641 'dot':170 'doublecircl':266 'draft':647 'drastic':426 'drop':887,916,1369 'edg':1264 'els':1538 'empir':368 'empti':500,550,1573 'end':158,1052,1371,1712 'engag':2 'english':1447 'enough':75,1455,1598 'equival':1459 'error/edge-case':1267 'evalu':1763 'everi':309,332,348,1693 'everyth':150,1537,1589 'evid':416,1130 'execut':725,932,937,962,1861 'exercis':1266 'exist':486,491,836,949,1785 'exit':1451 'experi':443,578,768,784,1144 'explain':385 'explicit':125,257,1414,1443 'explor':329 'extract':239,473,1357 'fact':1162 'fail':652,767,1227 'failur':414,1270 'fals':986 'fearless':321 'feat':995 'featur':775,1195 'feed':621 'feedback':594,1148,1287 'fewer':709 'ff4444':279 'ffffcc':254 'file':181,466,1092 'fill':252,277,1829 'fillcolor':253,278 'find':337,382,687,1583 'finish':1491,1843 'first':683,1122,1173,1229,1755 'five':56 'five-phas':55 'fix':13,694,901,996,1074,1218,1238,1244,1692,1753,1762,1774 'fix-typ':900 'fixm':575,1065,1683 'flag':1319,1568 'flaw':698,1060 'flight':1521 'focus':642,1376 'fold':752 'folklor':406 'follow':1525,1718 'follow-up':1524,1717 'fontcolor':280 'format':514 'found':1676 'frontier':215,328,797,805,875,1580,1701 'fundament':696 'gap':781,787,790,1354 'gate':218,881 'gave':1714 'gemini.md':469 'generat':705 'git':564,568,906,951,980,983,1787 'go':1421 'goal':475 'good':152,1597 'grace':1273 'grind':10,123,168 'ground':1120 'group':742 'guidanc':180,465 'halt':1399,1450 'handl':1032,1268,1272,1842 'happi':1627 'hasn':893 'hedg':363 'help':1290 'hesit':409 'heurist':184,498,547,626,1359,1366 'high':616 'higher':878,1704 'highest':676 'highest-leverag':675 'histori':206,592,907,952,1788 'human':1103 'hypothes':444 'ideal':615 'identifi':651,802 'immedi':1034,1432,1570,1647,1862 'implement':445,1235,1759,1767 'import':194,1261 'importance-r':193 'improv':12,102,858,1186,1237,1661,1761 'in-context':501 'in-flight':1519 'includ':1207 'incomplet':384,772 'incorpor':1038 'independ':938 'infer':487 'init':175,177,282,459 'initi':496 'inquiri':315 'insight':402 'inter':1379,1510,1557 'inter-cycl':1378,1509,1556 'interact':88 'interrupt':1013 'introduc':809,1112 'intuit':390 'invok':97 'iron':121 'issu':483,1054 'iter':436 'keep':48,101,717,1420 'keyword':1473 'known':482 'label':176,188,200,210,222,231,243,256,263,268,293,297,302 'lack':814,1470 'lag':613 'last':1142 'latest':584 'law':122 'lead':1334 'leader':322 'least':842,1708 'len':376 'let':1846 'letter':130 'level':388 'leverag':677 'lint/build':1134 'list':504,529,716,799,838,872,1572,1700 'll':1614,1741 'log':565,579,970 'logging/reporting':1293 'logic':972 'long':25,715 'long-run':24 'look':151,343,1560,1586 'loop':1329 'loss':1062 'lost':552 'low':618 'make':394 'mandatori':1171 'mani':1800 'march':431 'margin':703 'match':1765 'max':505,1367 'mean':1805 'measur':360,446,824,1326,1592 'mechan':399,404,928 'medium':617 'messag':994,1027,1634 'meta':1865 'meta-plan':1864 'methodolog':477 'metric':585,966,1139 'micro':737,1815 'micro-task':736,1814 'mid':1057,1489,1678 'mid-task':1056,1488 'mid-work':1677 'mindset':307 'miss':777 'ml/data':30 'ml/research':576,1138 'mode':8,93,1106,1110,1271 'model':110,1192 'modifi':1090 'movement':1140 'multipl':366 'must':728,845,1206 'n':516,1044,1386,1390,1500 'nannounc':246 'nassess':191 'nativ':926 'ndetect':179 'need':889,1236,1746,1760 'never':69,269,976,1656 'new':1531 'nexecut':225 'next':1011,1076,1204,1330,1374,1577,1668 'nground':234 'nheurist':238 'ninit':182 'nobodi':341 'non':271,689,856 'non-obvi':688,855 'none':485 'normal':1101 'note':611,896,1005 'noth':1806 'npriorit':213 'nreview':203 'nsignal':259 'nsolvabl':217 'nto':304 'observ':1291 'obvious':690,857,865,1282,1297,1576 'oldest':1370 'one':340,601,756,843,971,1358,1435,1438,1584,1709 'one-paragraph':600 'onelin':566 'open':587,593,654 'opencode.md':471 'oper':1099 'opportun':793 'order':758,936 'outcom':1145 'outlin':1848 'output':733,839,961,965,1189,1285,1673,1797,1808 'outsid':1094 'outweigh':699 'overrid':1175 'overview':46,59,187,190,283,284,296,522,539,555,1406,1431,1834 'overview/understand/plan/reflect':1507 'paper':645,1193 'paragraph':602 'pass':1242,1651,1653 'patch':866,1779 'path':826 'pattern':236,1312 'paus':65,242,244,289,1381,1419,1449,1512,1559,1618,1619 'per':461,943,974 'perf':1001,1595 'perform':1280 'performance/efficiency':792 'permiss':1105,1108 'persist':227,967,975 'phase':57,513,535,553,631,666,929,1029,1049,1079,1115,1506,1516,1687,1845,1854,1859 'phrase':41 'pioneer':325 'pivot':407,427 'plan':61,209,212,286,622,661,668,1069,1850,1858,1866 'platform':924 'point':1410 'polish':16,704 'polish/refinement':795 'polit':1460,1545 'prais':1542,1623 'prefer':518,1361 'prepend':520,1363 'primari':1188,1672 'principl':509,684 'print':1384,1495 'priorit':623,1690 'prioriti':757,800,935,1178 'problem':691,849,911,947,1220,1783 'proceed':628 'procrastin':1853 'produc':598,729 'progress':1250,1327,1616,1770,1826 'project':72,474,494,558,812,1086,1096,1585,1838 'prompt':990,1114 'properti':821 'provid':895 'qualiti':1301 'quality/coverage':780 'question':588,655,1172,1230,1555 'quick':525 'rankdir':173 'rate':195 'ration':1729,1730 're':541,1836 're-read':540,1835 'read':526,542,624,1837 'readme.md':472 'real':359,1219 'realiti':1731 'reason':681 'receiv':1149 'recent':649,905 'recogn':1446,1468,1474 'recommend':119 'red':1567 'redirect':1036 'refactor':1000 'reflect':63,230,233,288,512,1117,1165,1372,1383 'refresh':1332,1349 'regular':1528,1721 'reiniti':548 'relev':204,637 'reliabl':724 'remain':1019,1254,1427 'repeat':66 'replac':422 'reproduc':953,1789 'request':1465,1486,1639 'requir':1102 'research':31,357 'resolv':915,955,1791 'restart':1535,1725 'result':1133 'resum':1041,1643 'return':1346 'review':577,590,634,648,1147 'revis':591,596,1150 'right':109,679 'rigor':313 'rule':133,140 'run':26,570,582,873,1098 'ruthless':324 'safeti':1081 'satisfact':1628 'say':21,105 'scaffold':1200,1665 'scan':216,463,573,586,876,1253,1581,1702 'scenario':1262 'scientif':314 'scientist':320 'scope':712 'scratch':545 'second':1394 'section':778 'secur':1059,1295 'seem':1626 'self':1126,1159,1214 'self-assess':1125,1158 'self-audit':1213 'semant':992 'send':1441 'sens':395 'sent':1632 'session':27,99,183,311,462,497,546,625,1365,1532,1711,1727 'sever':1067,1684 'shape':185,197,207,219,228,240,249,260,265,274 'share':1021 'ship':1600 'sign':989 'signal':127,235,412,1121,1396,1415,1445,1566,1630,1659 'silenc':1539 'simplifi':1304 'singl':37,85,380 'single-angl':379 'skill':1191 'skill-autogrind' 'skip':1008,1156,1425 'sleep':1402,1428 'solid':665 'solut':862 'solvabl':880 'someth':810 'soon':1464,1554 'source-ttttonyhe' 'spirit':137 'stack':480 'stand':747,1820 'start':660,1388,1852 'state':192,353,543,559,605,1839 'state-of-the-art':352 'status':569,1135 'stay':1083 'steer':1050,1648 'step':722,1118,1166,1251,1307,1341,1355,1867 'still':948,1784 'stop':18,82,106,126,148,156,165,258,264,270,305,1395,1409,1433,1444,1448,1472,1487,1497,1502,1503,1513,1565,1621,1629,1649,1658,1742,1832 'strategi':419 'strong':118 'structur':370,490 'stubborn':430 'stuck':1328 'style':251,276,1002 'substant':755 'suggest':1546 'suit':572 'summar':1615 'summari':606 'support':942 'surfac':1299 'switch':1111 'symptom':1224 'synthes':377 'system':1091 'tabl':1343 'target':1670 'task':39,86,144,214,708,713,719,727,738,801,806,844,869,884,888,903,921,927,933,939,944,1012,1020,1058,1070,1212,1490,1494,1529,1540,1578,1694,1710,1720,1722,1775,1801,1816 'tb':174 'tech':479 'tempt':303 'ten':700 'termin':1485 'test':14,571,766,782,964,998,1132,1201 'test/validator':1749 'tests/tooling':1662 'tests/validations':1650 'theoret':369,397 'thing':451 'todo':574,837,853,1571 'todo/fixme':1699 'tool':112,1202 'topic-agent-skill' 'topic-agent-skills' 'topic-ai-agents' 'topic-autonomous-agents' 'topic-claude-code' 'topic-cursor' 'topic-gemini-cli' 'topic-llm-agent' 'topic-openai-codex' 'topic-opencode' 'topic-skills-sh' 'toward':432,437 'track':920 'train':581 'transfer':508 'treat':308 'trigger':40 'turbul':410 'two':387,1170 'type':902,1031 'unambigu':1483 'under':400 'understand':60,199,202,285,438,456,630,633,663 'unrel':978 'unresolv':1024 'unrestrict':111 'up':1526 'updat':420 'use':113,982,991 'user':78,892,1015,1026,1413,1440,1611,1622,1631,1713,1738 'ux':1284 'vagu':362 'valid':226,653,765,963,1226,1240,1247,1258 'valuabl':336 've':1603 'verifi':732,882,945,1129,1781,1813 'version':115 'violat':128,135 'visibl':731,1811 'vs':1141 'wait':247,1400,1424 'warn':267,301 'well':711 'well-scop':710 'white':281 'window':1828 'within':1084 'without':17,403,408,1243 'work':7,52,62,89,205,221,224,287,452,671,807,931,1197,1300,1591,1625,1679,1739 'workflow':458 'wouldn':745,1818 'wrap':1552 'write':34 'wrong':434,1751 'yes':294 'yet':818,823 '中文':1475,1484 '停':1476 '停止':1477 '够了':1479 '暂停':1478 '结束':1480 '自己动':43","prices":[{"id":"b898fe5a-55a4-433d-9d06-44adaa5aabf3","listingId":"a256a02a-5efe-41c8-bb5c-1789a0094f5d","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"ttttonyhe","category":"autogrind","install_from":"skills.sh"},"createdAt":"2026-05-18T13:20:54.101Z"}],"sources":[{"listingId":"a256a02a-5efe-41c8-bb5c-1789a0094f5d","source":"github","sourceId":"ttttonyhe/autogrind","sourceUrl":"https://github.com/ttttonyhe/autogrind","isPrimary":false,"firstSeenAt":"2026-05-18T13:20:54.101Z","lastSeenAt":"2026-05-18T19:13:32.793Z"}],"details":{"listingId":"a256a02a-5efe-41c8-bb5c-1789a0094f5d","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"ttttonyhe","slug":"autogrind","github":{"repo":"ttttonyhe/autogrind","stars":7,"topics":["agent-skill","agent-skills","ai-agents","autonomous-agents","claude-code","cursor","gemini-cli","llm-agent","openai-codex","opencode","skills-sh"],"license":"mit","html_url":"https://github.com/ttttonyhe/autogrind","pushed_at":"2026-04-12T06:44:43Z","description":"✊ Let your agent grind on your projects 24x7 fully autonomously","skill_md_sha":"82e4bb7c659fe6392c0b0c4ce4fe13d7032999ac","skill_md_path":"SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/ttttonyhe/autogrind"},"layout":"root","source":"github","category":"autogrind","frontmatter":{"name":"autogrind","license":"MIT","description":"Engage in 24x7 auto-work mode. Continuously grind through improvements, fixes, tests, and polish without stopping until you say so. For long-running sessions across code, ML/data, research, design, or writing. Not for single bounded tasks. Trigger phrases: /autogrind, /自己动, 'autogrind this', 'grind on this', 'keep working don't stop', 'work until I say stop', 'keep improving', 'keep going'.","compatibility":"Designed for Claude Code (or similar products)"},"skills_sh_url":"https://skills.sh/ttttonyhe/autogrind"},"updatedAt":"2026-05-18T19:13:32.793Z"}}