[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79XNT9TC7SFA8TP7RTT8E64":3,"skill-same:01M79XNT9TC7SFA8TP7RTT8E64":46},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M79XNT9TC7SFA8TP7RTT8E64","play-penta",null,"Start or join a Penta game and play an assigned seat through the session API. Handles local server startup, compact choices, batching, waiting and recovery against a human or another bot.","lacker\u002Fpenta","lacker","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F106475?v=4","01M79XNP7WPM09E70KSK02PMHA","Kevin Lacker","01M79XNP8X7SR6EHX6XX5G0WWS",1,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fplay-penta\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: play-penta\ndescription: Start or join a Penta game and play an assigned seat through the session API. Handles local server startup, compact choices, batching, waiting and recovery against a human or another bot.\n---\n\n# Play Penta\n\n## Native host route\n\nWhen given a native `penta-host` server URL and seat secret, use\n`node tools\u002Fpenta-agent\u002Fnative.mjs METHOD -` with JSON on stdin containing\n`server`, `secret`, and the method's arguments. Follow `docs\u002Fnative-matches.md`.\nThis route uses no hosted connection handle, shared agent daemon, or credential\nfiles. Do not run the hosted `up`\u002F`attach` commands for a native assignment.\n\nFor repeated play, use the maintained memory-only local session. Start\n`node tools\u002Fpenta-agent\u002Fsession-rpc.mjs serve -` with `{server, secret}` on\nstdin, then close stdin. Keep this process running and retain `clientServer`\nfrom its startup reply. For each request, run the finite command\n`node tools\u002Fpenta-agent\u002Fsession-rpc.mjs call -` with\n`{clientServer, secret, request: {method, ...args}}` on stdin. Use\n`exec_command` with an initial yield of 30000 ms; it returns early when the\nrequest command exits. Poll only while that finite process is still running.\nThis avoids persistent PTY collection delays while retaining session context.\nUse the local shell runtime that can reach the loopback host.\n\nRepeated rules use minimal rows containing `definition`, `reference`,\n`printed: null`, and `printedPreviouslySent: true`. The session also delivers\nnew seat-visible events using an explicit cursor; identical events remain\ndistinct. Unchanged ability metadata uses `{source,abilitiesPreviouslySent:true}`;\nnew, changed, or reappearing sources include full records. Every visible source\nremains listed. Current zones and pending decision options remain complete. After context compaction\nor lost tool output, send `{\"method\":\"observe\",\"full\":true}` to restore rules\nand event history. `sessionStats` returns calls, bytes, timings, and errors.\nSend `{\"method\":\"shutdown\"}` to close only your local session after play.\nThe JSON-lines `native.mjs session` route remains available for streaming\ncallers; avoid long fixed PTY output-collection intervals.\n\nFor normal native play, start with `{\"method\":\"next\"}` and then use one\n`interact` per decision. Native output defaults to `pilot-v3` with\n`position.actionInterface: \"proposal-v1\"`. Read complete seat-visible zones,\nprinted rules, `position.actionContext`, `position.actionAbilities`, and any\npending decision options. Construct your move; no exhaustive action menu is\nneeded. Supply `requestId`, current `revision`, and\n`choices: [{proposal: {action: YOUR_TYPED_ACTION, payment?: YOUR_PAYMENT}}]`.\nAnswer pending decisions with `{decision: ID, options: [IDs]}`; retain order.\nDecision option rows inherit `position.decision.options.defaults`; overlay each\nrow on those defaults. Every option ID and label remains inline. Mulligan actions\nare `KeepHand` and `TakeMulligan`; `next` accepts `maxWaitMs`, not `waitMs`.\n\nIn `first-to-two-wins`, follow `position.situation.match.stage` between games.\nFollow the private decision prompt. When asked for the next main deck, select\nall its option IDs, including retained cards; `Library` marks current main-deck copies and `OutsideGame` sideboard\ncopies. Your full main-deck and sideboard rules are available in the packet.\nAfter both submissions, answer the losing seat's play\u002Fdraw decision. Keep the\nsame host and seat secret, and stop only when the packet is `complete`.\n\nCast actions use `{type:\"CastSpell\",card,choices:{playOptionId,modeIds?,x?,\ntargetSelections?,alternativeCostId?,additionalCostIds?},costObjects?}`. Target\nselections use `{slotId,targets:[{type:\"player\",seat} or\n{type:\"permanent\"|\"card\"|\"spell\",objectId}],amounts?}`. Omitted arrays are empty,\nX is zero, and alternative cost is none. Do not add redundant flattened cast\nfields. Omitted `choices` uses those empty defaults, without choosing required\ntargets or modes. Ability actions use `source`, the provided `ability` identity, and\noptional `targetSelections`, `costObjects`, `modeIds`, and `x`. For selected\nmodes, concatenate their target slots after the base ability slots in selection\norder and assign sequential slot IDs; mode-local IDs are not combined IDs.\nCast `costObjects` names resources sacrificed, discarded, or exiled for costs\n(including collect evidence), separately from targets. Inspect an object's\n`grantedCastCosts` for runtime alternative-cost IDs such as granted flashback.\nIts `playPermission` describes an applicable exile\u002Fgraveyard permission and cost.\nAirbend's `alternativeMana` cost applies with the default\u002Fnull alternative-cost\nID; it is not an entry in `grantedCastCosts`. Timing and final payment still apply.\n\nOmitting payment delegates payment planning for your chosen action. When mana\nsources matter, supply `payment:{funding:[{action:ACTIVATE_MANA_ABILITY,\nmanaUnits?:[IDs],answers?:[[OPTION_IDS]]}],announcements?:[[OPTION_IDS]],\nmanaUnits?:[IDs],contributions?:[{source,kind,symbol}]}`. Funding is ordered;\nexplicit payment does not silently select more sources. Use\n`combination:{blue:1,green:1}` instead of a separate color when splitting mana\nfrom one ability; otherwise supply `color`. Contributions use\n`convoke`, `delve`, `improvise`, or `waterbend`. Read `docs\u002Fnative-matches.md`\nand `node tools\u002Fpenta-agent\u002Fnative.mjs schema` for the complete typed contract;\n`tools\u002Fpenta-agent\u002Fpilot-v3.d.ts` contains generated types.\n\nAn optional `yield:{until:[...],maxPasses:64}` records your explicit decision to\npass those windows. Empty choices with a yield are supported. The session\nsubmits once and renews read-only waits up to 60 seconds (`maxWaitMs`, local\nonly), retaining the original receipt. Read `host.receipt.yieldInstallation`:\na partial batch or new own decision prevents installation. Use `next` after a\nwait timeout; never resubmit a committed move merely to wait. A rejected\nproposal reports `host.receipt.rejection` with code, path and message; a batch\ncan have a committed prefix. Retain the exact mutation until its outcome is\nknown and retry identical arguments after uncertain output, or query `receipt`.\nCombat declarations can auto-finish after the last eligible attacker\u002Fblocker.\nAvoid blindly appending a finish action; a redundant finish rejects only that\nsuffix and prevents installing its yield. Inspect the returned action context.\n\nUse `host.currentYield` for current policy and `yieldAtAcceptance` for the\nacceptance snapshot. Read `host.receipt.updates` alongside `position.updates`;\nwaiting events appear in `host.eventDelivery.items`. After context loss call\n`observe` with `full:true` to restore printed rules and events. Use `inspect_ref`\nfor referenced details. No per-move explanation is needed.\n\nIf you need optional legal-action discovery, explicitly call `discover` with\n`presentation:\"pilot-v2\"`, then use `choose` with the displayed ticket and a\nnew `requestId`, or copy the exact action into `{action:...}`. Legacy menu views\nare opt-in; normal native play does not discover or paginate exhaustive menus.\n\nFor \"pass until phase X unless the opponent acts\", call `yield` with the current\n`revision`, a new `requestId`, and an OR list such as\n`until: [{type:\"nextStep\",step:\"End\"},{type:\"opponentActs\"}]`.\nUse `PrecombatMain`, `BeginningOfCombat`, or `PostcombatMain` for those phase\nboundaries. `nextStep` means a future entry, not the current occurrence;\n`activeSeat` optionally selects `self`, `opponent`, or `either`.\n`opponentActs` includes mana activations. Use `opponentCastsSpell` for a narrower\ninterrupt, or add `nonManaAction` to stop when another legal action is available.\nThis is an explicit decision to pass the covered priority windows, not a change\nto which actions are forced. Non-forced decisions still need the pilot.\nRead `host.currentYield` for the stop reason; `cancelYield` with a new `requestId` cancels\nthe instruction. Yields expire at game boundaries and the requested pass limit\n(`maxPasses`, default 64, maximum 256). Conditions and updates survive reconnects\nwhile the host remains alive.\n\nAfter model context loss, reuse the same seat secret, resolve any uncertain\nmove, call `observe`, and recover seat-visible `transcript` pages and notes.\nKeep the host alive: state is volatile and a host crash loses the live match.\nOnly the organizer obtains both seats, exports private debug data, or closes\nthe host. Never inspect the other pilot's credentials, transcript or process.\nThis workflow is not a hardened sandbox for agents sharing unrestricted shell\naccess. When asked to start a native match, follow the native guide's binary\nstartup and distribute only each pilot's own credential.\nStart each pilot with fresh context, not inherited organizer history. With\nCodex subagents, use `fork_turns: \"none\"` and a brief containing the skill,\nserver URL, assigned seat, and only that seat's secret.\n\n## Hosted route\n\nThe instructions below apply only to hosted room\u002Fconnection assignments. Native\nserver\u002Fseat-secret assignments use the route above, including its batch shapes.\n\nUse `node tools\u002Fpenta-agent\u002Fcli.mjs COMMAND 'JSON'` from the repository root.\nArguments can also come from `@FILE` or `-` for stdin; use those for credentials.\nThe maintained client handles transport and presentation. All gameplay choices\nremain yours. No MCP configuration, client restart or model API key is needed.\nUse `--help` for command arguments and `docs\u002Fbot-sessions.md` for exact forms.\n\n## Start or join\n\nFor a new local game, run `up`, then `options` to discover formats and decks.\n`up` installs missing web packages and starts the existing web\u002FWASM development\nserver. If it returns `starting`, call it again after inspecting the returned\nlog as needed; do not start a second build. Diagnose missing prerequisites from\nthat log and `web\u002FREADME.md`. `PENTA_SERVER_URL` selects an existing deployment\nand disables local server startup. Keep the same environment for every command.\n\nCreate a game with `start_match` and a unique `requestId`, `format`, `p1Deck`,\nand `p2Deck`. Set `humanSeat` to `p1` or `p2` for a human opponent and open the\nreturned `humanUrl`. `matchMode` defaults to `first-to-two-wins`; use\n`one-conclusion` when requested. Repeating the identical creation request\nretrieves its saved result; do not invent another ID to recover lost output.\n\nThe organizer receives both seat credentials. Give each player only its own\n`{room, token}`. When assigned an existing seat, skip creation and attach only\nto that seat. `attach` defaults to `presentation: \"decision-v1\"`. Retain the\nreturned `connection`. The local session process persists across CLI commands;\n`status` lists connection handles after context loss. Do not read its private\nrecovery files to inspect another player's credentials or game information.\n\n## Play\n\nAt `ready`, read the current position, ordered updates, and available choices.\nEach `{shared, rows}` table is self-contained: shared fields apply to every row.\nIDs distinguish objects and option IDs are not array positions. Printed card\nreferences can differ from current observed characteristics after effects.\n\nUse `choose '{\"ticket\":\"...\"}'` for a concrete action. Decision tickets require\nan explicit `options` array, including `[]` for an empty selection. Preserve\noption IDs and their order. The engine advances truly forced continuations;\nall remaining gameplay choices are yours. No per-move explanation is required.\n\nSubmit an already-decided group with `play`, supplying `connection`, `revision`,\nand `choices: [{ticket: \"...\"}, {ticket: \"...\"}]`. Tickets must belong to the\ncurrent view. Batch multiple attackers or block assignments together. Exact\nactions may be mixed in, including `{action: {type: \"FinishDeclaringAttackers\"}}`\nor `FinishDeclaringBlockers` when you choose to commit the declaration. The\nengine may already have advanced a forced finish: inspect `receipt.accepted`\nand the returned state. Accepted prefixes stay committed.\n\nA `BottomCards` ticket includes the whole mulligan-bottom group. Submit all\nselected option IDs for one decision together. Split a batch when new\ninformation or an unresolved choice is needed; batching never licenses making\nthat choice automatically. `play` and `choose` return the next view: use it\ndirectly when ready instead of calling `next` again.\n\nUse `inspect_ref` with `{reference, offset}` for omitted details; follow\n`nextOffset` until the needed entries are available. References are frozen to\ntheir view. Use `next` with `{connection}` when waiting, and\n`{connection, full: true}` for context refresh or expired references. Calls wait\nup to 25 seconds. Full refreshes resend card text within the inline budget;\nremaining text is inspectable. Stop at `complete`.\n\n## Recover\n\nAfter an uncertain submission or lost CLI result, use `retry` with\n`{connection}`. It retains the exact resolved body and request ID, including\nfor a batch or a successful move whose output was lost. Repeating an identical\nretained `choose` ticket and options also recovers its receipt. Resolve the\nuncertainty before another move; do not change the submitted options.\nAfter a definite stale refusal, refresh and choose again from the new view.\n\nIf the local session process restarts, saved connections and retry bodies\nsurvive. Recover any outstanding move with `retry`, then refresh with\n`next(full: true)` because ordinary tickets and references expire on restart.\nAn uncertain room creation whose credentials never reached the client is\nreported explicitly; stop and explain it rather than silently dealing again.\n\n`stop` ends the session process and leaves the game server running. Ordinary\nplay does not require it. Report engine or information discrepancies with\nvisible position, revision, ticket and receipt context; omit seat credentials.\n","# Play Penta\n\n## Native host route\n\nWhen given a native `penta-host` server URL and seat secret, use\n`node tools\u002Fpenta-agent\u002Fnative.mjs METHOD -` with JSON on stdin containing\n`server`, `secret`, and the method's arguments. Follow `docs\u002Fnative-matches.md`.\nThis route uses no hosted connection handle, shared agent daemon, or credential\nfiles. Do not run the hosted `up`\u002F`attach` commands for a native assignment.\n\nFor repeated play, use the maintained memory-only local session. Start\n`node tools\u002Fpenta-agent\u002Fsession-rpc.mjs serve -` with `{server, secret}` on\nstdin, then close stdin. Keep this process running and retain `clientServer`\nfrom its startup reply. For each request, run the finite command\n`node tools\u002Fpenta-agent\u002Fsession-rpc.mjs call -` with\n`{clientServer, secret, request: {method, ...args}}` on stdin. Use\n`exec_command` with an initial yield of 30000 ms; it returns early when the\nrequest command exits. Poll only while that finite process is still running.\nThis avoids persistent PTY collection delays while retaining session context.\nUse the local shell runtime that can reach the loopback host.\n\nRepeated rules use minimal rows containing `definition`, `reference`,\n`printed: null`, and `printedPreviouslySent: true`. The session also delivers\nnew seat-visible events using an explicit cursor; identical events remain\ndistinct. Unchanged ability metadata uses `{source,abilitiesPreviouslySent:true}`;\nnew, changed, or reappearing sources include full records. Every visible source\nremains listed. Current zones and pending decision options remain complete. After context compaction\nor lost tool output, send `{\"method\":\"observe\",\"full\":true}` to restore rules\nand event history. `sessionStats` returns calls, bytes, timings, and errors.\nSend `{\"method\":\"shutdown\"}` to close only your local session after play.\nThe JSON-lines `native.mjs session` route remains available for streaming\ncallers; avoid long fixed PTY output-collection intervals.\n\nFor normal native play, start with `{\"method\":\"next\"}` and then use one\n`interact` per decision. Native output defaults to `pilot-v3` with\n`position.actionInterface: \"proposal-v1\"`. Read complete seat-visible zones,\nprinted rules, `position.actionContext`, `position.actionAbilities`, and any\npending decision options. Construct your move; no exhaustive action menu is\nneeded. Supply `requestId`, current `revision`, and\n`choices: [{proposal: {action: YOUR_TYPED_ACTION, payment?: YOUR_PAYMENT}}]`.\nAnswer pending decisions with `{decision: ID, options: [IDs]}`; retain order.\nDecision option rows inherit `position.decision.options.defaults`; overlay each\nrow on those defaults. Every option ID and label remains inline. Mulligan actions\nare `KeepHand` and `TakeMulligan`; `next` accepts `maxWaitMs`, not `waitMs`.\n\nIn `first-to-two-wins`, follow `position.situation.match.stage` between games.\nFollow the private decision prompt. When asked for the next main deck, select\nall its option IDs, including retained cards; `Library` marks current main-deck copies and `OutsideGame` sideboard\ncopies. Your full main-deck and sideboard rules are available in the packet.\nAfter both submissions, answer the losing seat's play\u002Fdraw decision. Keep the\nsame host and seat secret, and stop only when the packet is `complete`.\n\nCast actions use `{type:\"CastSpell\",card,choices:{playOptionId,modeIds?,x?,\ntargetSelections?,alternativeCostId?,additionalCostIds?},costObjects?}`. Target\nselections use `{slotId,targets:[{type:\"player\",seat} or\n{type:\"permanent\"|\"card\"|\"spell\",objectId}],amounts?}`. Omitted arrays are empty,\nX is zero, and alternative cost is none. Do not add redundant flattened cast\nfields. Omitted `choices` uses those empty defaults, without choosing required\ntargets or modes. Ability actions use `source`, the provided `ability` identity, and\noptional `targetSelections`, `costObjects`, `modeIds`, and `x`. For selected\nmodes, concatenate their target slots after the base ability slots in selection\norder and assign sequential slot IDs; mode-local IDs are not combined IDs.\nCast `costObjects` names resources sacrificed, discarded, or exiled for costs\n(including collect evidence), separately from targets. Inspect an object's\n`grantedCastCosts` for runtime alternative-cost IDs such as granted flashback.\nIts `playPermission` describes an applicable exile\u002Fgraveyard permission and cost.\nAirbend's `alternativeMana` cost applies with the default\u002Fnull alternative-cost\nID; it is not an entry in `grantedCastCosts`. Timing and final payment still apply.\n\nOmitting payment delegates payment planning for your chosen action. When mana\nsources matter, supply `payment:{funding:[{action:ACTIVATE_MANA_ABILITY,\nmanaUnits?:[IDs],answers?:[[OPTION_IDS]]}],announcements?:[[OPTION_IDS]],\nmanaUnits?:[IDs],contributions?:[{source,kind,symbol}]}`. Funding is ordered;\nexplicit payment does not silently select more sources. Use\n`combination:{blue:1,green:1}` instead of a separate color when splitting mana\nfrom one ability; otherwise supply `color`. Contributions use\n`convoke`, `delve`, `improvise`, or `waterbend`. Read `docs\u002Fnative-matches.md`\nand `node tools\u002Fpenta-agent\u002Fnative.mjs schema` for the complete typed contract;\n`tools\u002Fpenta-agent\u002Fpilot-v3.d.ts` contains generated types.\n\nAn optional `yield:{until:[...],maxPasses:64}` records your explicit decision to\npass those windows. Empty choices with a yield are supported. The session\nsubmits once and renews read-only waits up to 60 seconds (`maxWaitMs`, local\nonly), retaining the original receipt. Read `host.receipt.yieldInstallation`:\na partial batch or new own decision prevents installation. Use `next` after a\nwait timeout; never resubmit a committed move merely to wait. A rejected\nproposal reports `host.receipt.rejection` with code, path and message; a batch\ncan have a committed prefix. Retain the exact mutation until its outcome is\nknown and retry identical arguments after uncertain output, or query `receipt`.\nCombat declarations can auto-finish after the last eligible attacker\u002Fblocker.\nAvoid blindly appending a finish action; a redundant finish rejects only that\nsuffix and prevents installing its yield. Inspect the returned action context.\n\nUse `host.currentYield` for current policy and `yieldAtAcceptance` for the\nacceptance snapshot. Read `host.receipt.updates` alongside `position.updates`;\nwaiting events appear in `host.eventDelivery.items`. After context loss call\n`observe` with `full:true` to restore printed rules and events. Use `inspect_ref`\nfor referenced details. No per-move explanation is needed.\n\nIf you need optional legal-action discovery, explicitly call `discover` with\n`presentation:\"pilot-v2\"`, then use `choose` with the displayed ticket and a\nnew `requestId`, or copy the exact action into `{action:...}`. Legacy menu views\nare opt-in; normal native play does not discover or paginate exhaustive menus.\n\nFor \"pass until phase X unless the opponent acts\", call `yield` with the current\n`revision`, a new `requestId`, and an OR list such as\n`until: [{type:\"nextStep\",step:\"End\"},{type:\"opponentActs\"}]`.\nUse `PrecombatMain`, `BeginningOfCombat`, or `PostcombatMain` for those phase\nboundaries. `nextStep` means a future entry, not the current occurrence;\n`activeSeat` optionally selects `self`, `opponent`, or `either`.\n`opponentActs` includes mana activations. Use `opponentCastsSpell` for a narrower\ninterrupt, or add `nonManaAction` to stop when another legal action is available.\nThis is an explicit decision to pass the covered priority windows, not a change\nto which actions are forced. Non-forced decisions still need the pilot.\nRead `host.currentYield` for the stop reason; `cancelYield` with a new `requestId` cancels\nthe instruction. Yields expire at game boundaries and the requested pass limit\n(`maxPasses`, default 64, maximum 256). Conditions and updates survive reconnects\nwhile the host remains alive.\n\nAfter model context loss, reuse the same seat secret, resolve any uncertain\nmove, call `observe`, and recover seat-visible `transcript` pages and notes.\nKeep the host alive: state is volatile and a host crash loses the live match.\nOnly the organizer obtains both seats, exports private debug data, or closes\nthe host. Never inspect the other pilot's credentials, transcript or process.\nThis workflow is not a hardened sandbox for agents sharing unrestricted shell\naccess. When asked to start a native match, follow the native guide's binary\nstartup and distribute only each pilot's own credential.\nStart each pilot with fresh context, not inherited organizer history. With\nCodex subagents, use `fork_turns: \"none\"` and a brief containing the skill,\nserver URL, assigned seat, and only that seat's secret.\n\n## Hosted route\n\nThe instructions below apply only to hosted room\u002Fconnection assignments. Native\nserver\u002Fseat-secret assignments use the route above, including its batch shapes.\n\nUse `node tools\u002Fpenta-agent\u002Fcli.mjs COMMAND 'JSON'` from the repository root.\nArguments can also come from `@FILE` or `-` for stdin; use those for credentials.\nThe maintained client handles transport and presentation. All gameplay choices\nremain yours. No MCP configuration, client restart or model API key is needed.\nUse `--help` for command arguments and `docs\u002Fbot-sessions.md` for exact forms.\n\n## Start or join\n\nFor a new local game, run `up`, then `options` to discover formats and decks.\n`up` installs missing web packages and starts the existing web\u002FWASM development\nserver. If it returns `starting`, call it again after inspecting the returned\nlog as needed; do not start a second build. Diagnose missing prerequisites from\nthat log and `web\u002FREADME.md`. `PENTA_SERVER_URL` selects an existing deployment\nand disables local server startup. Keep the same environment for every command.\n\nCreate a game with `start_match` and a unique `requestId`, `format`, `p1Deck`,\nand `p2Deck`. Set `humanSeat` to `p1` or `p2` for a human opponent and open the\nreturned `humanUrl`. `matchMode` defaults to `first-to-two-wins`; use\n`one-conclusion` when requested. Repeating the identical creation request\nretrieves its saved result; do not invent another ID to recover lost output.\n\nThe organizer receives both seat credentials. Give each player only its own\n`{room, token}`. When assigned an existing seat, skip creation and attach only\nto that seat. `attach` defaults to `presentation: \"decision-v1\"`. Retain the\nreturned `connection`. The local session process persists across CLI commands;\n`status` lists connection handles after context loss. Do not read its private\nrecovery files to inspect another player's credentials or game information.\n\n## Play\n\nAt `ready`, read the current position, ordered updates, and available choices.\nEach `{shared, rows}` table is self-contained: shared fields apply to every row.\nIDs distinguish objects and option IDs are not array positions. Printed card\nreferences can differ from current observed characteristics after effects.\n\nUse `choose '{\"ticket\":\"...\"}'` for a concrete action. Decision tickets require\nan explicit `options` array, including `[]` for an empty selection. Preserve\noption IDs and their order. The engine advances truly forced continuations;\nall remaining gameplay choices are yours. No per-move explanation is required.\n\nSubmit an already-decided group with `play`, supplying `connection`, `revision`,\nand `choices: [{ticket: \"...\"}, {ticket: \"...\"}]`. Tickets must belong to the\ncurrent view. Batch multiple attackers or block assignments together. Exact\nactions may be mixed in, including `{action: {type: \"FinishDeclaringAttackers\"}}`\nor `FinishDeclaringBlockers` when you choose to commit the declaration. The\nengine may already have advanced a forced finish: inspect `receipt.accepted`\nand the returned state. Accepted prefixes stay committed.\n\nA `BottomCards` ticket includes the whole mulligan-bottom group. Submit all\nselected option IDs for one decision together. Split a batch when new\ninformation or an unresolved choice is needed; batching never licenses making\nthat choice automatically. `play` and `choose` return the next view: use it\ndirectly when ready instead of calling `next` again.\n\nUse `inspect_ref` with `{reference, offset}` for omitted details; follow\n`nextOffset` until the needed entries are available. References are frozen to\ntheir view. Use `next` with `{connection}` when waiting, and\n`{connection, full: true}` for context refresh or expired references. Calls wait\nup to 25 seconds. Full refreshes resend card text within the inline budget;\nremaining text is inspectable. Stop at `complete`.\n\n## Recover\n\nAfter an uncertain submission or lost CLI result, use `retry` with\n`{connection}`. It retains the exact resolved body and request ID, including\nfor a batch or a successful move whose output was lost. Repeating an identical\nretained `choose` ticket and options also recovers its receipt. Resolve the\nuncertainty before another move; do not change the submitted options.\nAfter a definite stale refusal, refresh and choose again from the new view.\n\nIf the local session process restarts, saved connections and retry bodies\nsurvive. Recover any outstanding move with `retry`, then refresh with\n`next(full: true)` because ordinary tickets and references expire on restart.\nAn uncertain room creation whose credentials never reached the client is\nreported explicitly; stop and explain it rather than silently dealing again.\n\n`stop` ends the session process and leaves the game server running. Ordinary\nplay does not require it. Report engine or information discrepancies with\nvisible position, revision, ticket and receipt context; omit seat credentials.\n",{"name":5,"description":7},"main","581e7ce37cc1113b0cec4873da4e089b0b8ccd6d",".agents\u002Fskills\u002Fplay-penta",3,"https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta","https:\u002F\u002Fgithub.com\u002Flacker","lacker.io","2026-11-10 01:12:38",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add lacker\u002Fpenta --skill play-penta",{"tool":38,"command":39},"bunx","bunx skills add lacker\u002Fpenta --skill play-penta",{"tool":41,"command":42},"pnpm","pnpm dlx skills add lacker\u002Fpenta --skill play-penta","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fplay-penta\u002FSKILL.md",true,[],[47,53,59,65,70,76],{"skillNo":48,"name":49,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M79XPSP16J8YZXH2NMGKD13S","refresh-magic-references","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frefresh-magic-references\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M79XPREGKGKZ5SK4G3PRVMZX","query-magic-references","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fquery-magic-references\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":19},"01M79XPPYT418SK0PCT0THY2DC","profile-engine-performance","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fprofile-engine-performance\u002FSKILL.md",[],[],{"skillNo":66,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M79XPMR0JQV6DPMWMQWQ95E3","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fplay-penta\u002FSKILL.md",[],[],{"skillNo":71,"name":49,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M79XPD0V6NC9EGGFF32QT6FJ","Locate, inspect, migrate, refresh, and index this repository's optional shared Magic: The Gathering reference cache, including the Wizards Comprehensive Rules, Eternal Central 93\u002F94 rules, Scryfall bulk card and ruling data, and schema-versioned SQLite indexes. Use when an agent needs reproducible local rules or card data, needs to diagnose cache freshness or locking, needs to build or repair the Scryfall index, or is asked to populate or update the clone-wide cache. Do not require it for questions that authoritative online sources can answer reliably.","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frefresh-magic-references\u002FSKILL.md",[],[],{"skillNo":77,"name":55,"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},"01M79XP5947747YQEYWE6BT39F","Query this repository's optional generated Scryfall SQLite index for exact printings, set and collector metadata, card characteristics, faces, Oracle text, legalities, keywords, colors, related parts, and rulings. Use for fast UUID or name lookups, full-text searches, joins, aggregate analysis, batch audits, or comparisons while developing Penta. Do not use it as the authority for Comprehensive Rules, live prices, or printing languages outside the indexed sources; refresh the cache or use an authoritative online source when necessary.","https:\u002F\u002Fgithub.com\u002Flacker\u002Fpenta\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fquery-magic-references\u002FSKILL.md",[],[]]