[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CDXNJDX2A54FAVWR41XMCC":3,"skill-same:01M7CDXNJDX2A54FAVWR41XMCC":45},{"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":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7CDXNJDX2A54FAVWR41XMCC","plan",null,"Use to plan or repeatedly validate a ticket's implementation RFC against its governing docs and code before implementation. Use when a ticket needs an RFC, or an existing RFC needs another validation pass.","dveyarangi\u002Fgoodwolf-harness","dveyarangi","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F693404?v=4","01M77KAGWZ8X0CNA092ZDNEF6T","Fima Rotter","01M7CDVSGB58A879MVTYC1SVD0",0,"MIT","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fplan\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: plan\ndescription: >-\n  Use to plan or repeatedly validate a ticket's implementation RFC against\n  its governing docs and code before implementation. Use when a ticket\n  needs an RFC, or an existing RFC needs another validation pass.\n---\n\n\u003Cstraw-dog question=\"q-oehis2\">Mechanism: not yet\u003C\u002Fstraw-dog>\n\n- Explore documentation in depth, follow links in it to find all decisions\n  relevant to current task. Find out which documented boundaries are involved,\n  and whether implementation challenges them.\n\n- If the ticket at hand is too coarse to yield a single unambiguous RFC, stop\n  planning and decompose it into sub-tickets via\n  [\u002Fticket](..\u002Fticket\u002FSKILL.md) first; then plan the first child.\n\n- An RFC implements a ticket — no ticketless RFCs. If the task at hand has no\n  ticket, create it first per `\u002Fticket`; its behavior-altitude rule governs\n  the criteria (shape stays in the RFC).\n\n- When incepting, ask yourself - what is the cheapest real example that could\n  prove this design assumption wrong?\n\n- Run [\u002Fimpact](..\u002Fimpact\u002FSKILL.md) on the planned change. Record its output on\n  the RFC.\n\n- Establish implementation scope, pinpoint and mention affected boundaries,\n  contracts, ownership. Present main code shapes - especially ones that\n  represent documented boundaries. Present code flows.\n\n- Detect implementation challenges and use [\u002Falign](..\u002Falign\u002FSKILL.md) to\n  resolve them with the user.\n\n- Do not select implementation shapes just because \"that how it is usually\n  done\" or based on first idea. Promote simplicity, look for elegant solutions,\n  prefer removing over expanding, prefer conciseness over verbosity.\n\n- A hidden edge the plan names — a race, a migration step, a special case — that a\n  choice of name, place or form brings with it reopens that choice before anything\n  is built around it.\n\n- Do not invent impl nouns or verbs that are not discussed or approved by\n  user. Use existing vocabulary.\n\n- If you see a concern, first check deeper how the existing architecture\n  documentation describes it - it most probably already does. Read\n  architecture.md, ADRs and the question store for this.\n\n- Major goal of this planning is to find inconsistencies in the pre-planned\n  architecture. Do this diligently. If such inconsistency found, do not stick\n  blindly for architectural decision - instead raise concern with user to\n  resolve it - either in code or in arch docs.\n\n- Make the RFC determinate where a choice shapes observable behavior, a\n  boundary or interface, ownership, failure semantics, compatibility,\n  migration, or another non-local constraint. State the shaping fact with a\n  reference to its durable owner; never rely on session context. Leave\n  reversible implementation-local choices to `\u002Ftdd` and `\u002Fimplement` unless\n  they become load-bearing.\n\n- The RFC (original or amended) must not describe load-bearing architecture\n  absent from the architecture docs — land the decision in the docs first\n  (`\u002Falign` when needed), then reference it from the RFC. What fails the entry\n  file's load-bearing counter-test belongs in code and its comments, however\n  closely an architecture sentence already describes it. Write each sentence\n  landed there as its invariant and its pointer: the reasons stay with the\n  decision that holds them, and what a mechanism or a script owns stays there.\n\n- Describe the implementation stages; allocate them according to `\u002Ftdd`\n  rules. Look at stages to make sure each of them keeps the tests green; in\n  case that would take too much temporary effort, allow red, note about test\n  status and compact reason for it.\n\n- Map out scope-specific limitations, follow-ups and related out-of-scope\n  concerns.\n\n- Prefer reuse and reduction of existing nouns, verbs and adjectives; when\n  adding shape - make sure documentation supports it, if in any doubt - `\u002Falign`\n  with user.\n\n- Any wrinkle against architecture should be resolved; when solution is not\n  found or following first-principles makes the code weird - align with user.\n  Architecture must lead code shape even if code disagrees. Do not expose\n  public methods or create flows that are not in architecture.\n\n- Each choice the RFC takes names the line of the governing question's accepted\n  `Where it stands` it keeps; a choice that keeps none, or crosses one, goes to\n  `\u002Falign` before it is planned.\n\n- If planned code includes a temporary solution that is dissolved by future\n  development, add RFC instruction to append TODO comment to code that flags\n  this. Otherwise code changes can start relying or considering the temporal\n  code, which can be hard to disentangle later.\n\n- Repeated invocations are validation passes over the same RFC. Re-read\n  the ticket, RFC, durable docs, and relevant code; challenge the plan against\n  new evidence and amend it in place. Do not create a replacement RFC merely\n  because the plan changed before implementation.\n\n- As an additional pass, try to explain things to yourself simply, as if you\n  are teaching the architecture, and being asked reasonable question and look\n  for areas that evade simple or common-sense explanation.\n\n- Validate the RFC adversarially on every pass:\n  - Probe each boundary and stage with counterexamples, including empty, partial,\n    faulting, and raced outcomes where relevant.\n  - Make every planned test prove the intended behavior and failure reason, not\n    merely that the path succeeds or raises.\n  - Separate sourced facts and existing invariants from assumptions; verify\n    facts in code or authoritative sources and surface load-bearing assumptions\n    for `\u002Falign`.\n  - Reject pseudocode or prescribed structure that introduces undocumented\n    architecture or freezes a reversible local choice.\n  - Check that every new promise names how it will be validated, and that\n    the stages collectively prove the ticket's acceptance criteria.\n\n- Repeat validation using this skill until no load-bearing ambiguity,\n  contradiction, or unverified assumption remains. The RFC may permit multiple\n  equivalent local implementations when they preserve the same documented shape\n  and proof.\n\n- If repeated validation keep bringing up gaps, take a step back and look\n  at what causes this oscillation. The architecture, ADRs, tickets are not\n  carved in stone, they can have real contradictions or inconsistencies that cause\n  it. Look into the core reasons and `\u002Falign` on them again if needed.\n\n- Overall, always consider future development and potential code reuse when\n  selecting code shapes.\n\n- Cover migrations, compatibility, rollout, failure handling, and observability\n  when relevant.\n\n- Record the plan as an RFC; its name and home are installed below.\n\n- RFC header: H1 is the ticket's title plus ` — implementation plan`; then\n  `- **record of** what it intends to become`, which the mover turns to\n  `what happened` at close; then `**Authored:** YYYY-MM-DD`, and `**Last amended:** YYYY-MM-DD` once a later\n  pass changes the plan; then one line stating what it implements, linking the\n  ticket.\n\n- DO NOT CHANGE THE CODE!\n\n## Installed from other mechanisms\n\n\u003Cinstalled by=\"ticket\">\n**P6** Name the RFC by the owning ticket's basename, under `docs\u002Frfc\u002F`, with no serial of its own; it\nmoves to `docs\u002Frfc\u002Fdone\u002F` with the ticket.\n\u003C\u002Finstalled>\n\n\u003Cinstalled by=\"session\">\n**S7** At the step's start, hold the unit by the question its work answers, `sessions.py --hold \u003Cq-id>\n--session \u003Ctag>`; a unit this session held is let go by it: say which, and that it is nobody's\nnow. A refusal naming another running session stops the step: say whose the unit is, and do\nnothing on it. A refusal saying the question is no unit leaves the step to go on, holding\nnothing.\n\u003C\u002Finstalled>\n","\u003Cstraw-dog question=\"q-oehis2\">Mechanism: not yet\u003C\u002Fstraw-dog>\n\n- Explore documentation in depth, follow links in it to find all decisions\n  relevant to current task. Find out which documented boundaries are involved,\n  and whether implementation challenges them.\n\n- If the ticket at hand is too coarse to yield a single unambiguous RFC, stop\n  planning and decompose it into sub-tickets via\n  [\u002Fticket](..\u002Fticket\u002FSKILL.md) first; then plan the first child.\n\n- An RFC implements a ticket — no ticketless RFCs. If the task at hand has no\n  ticket, create it first per `\u002Fticket`; its behavior-altitude rule governs\n  the criteria (shape stays in the RFC).\n\n- When incepting, ask yourself - what is the cheapest real example that could\n  prove this design assumption wrong?\n\n- Run [\u002Fimpact](..\u002Fimpact\u002FSKILL.md) on the planned change. Record its output on\n  the RFC.\n\n- Establish implementation scope, pinpoint and mention affected boundaries,\n  contracts, ownership. Present main code shapes - especially ones that\n  represent documented boundaries. Present code flows.\n\n- Detect implementation challenges and use [\u002Falign](..\u002Falign\u002FSKILL.md) to\n  resolve them with the user.\n\n- Do not select implementation shapes just because \"that how it is usually\n  done\" or based on first idea. Promote simplicity, look for elegant solutions,\n  prefer removing over expanding, prefer conciseness over verbosity.\n\n- A hidden edge the plan names — a race, a migration step, a special case — that a\n  choice of name, place or form brings with it reopens that choice before anything\n  is built around it.\n\n- Do not invent impl nouns or verbs that are not discussed or approved by\n  user. Use existing vocabulary.\n\n- If you see a concern, first check deeper how the existing architecture\n  documentation describes it - it most probably already does. Read\n  architecture.md, ADRs and the question store for this.\n\n- Major goal of this planning is to find inconsistencies in the pre-planned\n  architecture. Do this diligently. If such inconsistency found, do not stick\n  blindly for architectural decision - instead raise concern with user to\n  resolve it - either in code or in arch docs.\n\n- Make the RFC determinate where a choice shapes observable behavior, a\n  boundary or interface, ownership, failure semantics, compatibility,\n  migration, or another non-local constraint. State the shaping fact with a\n  reference to its durable owner; never rely on session context. Leave\n  reversible implementation-local choices to `\u002Ftdd` and `\u002Fimplement` unless\n  they become load-bearing.\n\n- The RFC (original or amended) must not describe load-bearing architecture\n  absent from the architecture docs — land the decision in the docs first\n  (`\u002Falign` when needed), then reference it from the RFC. What fails the entry\n  file's load-bearing counter-test belongs in code and its comments, however\n  closely an architecture sentence already describes it. Write each sentence\n  landed there as its invariant and its pointer: the reasons stay with the\n  decision that holds them, and what a mechanism or a script owns stays there.\n\n- Describe the implementation stages; allocate them according to `\u002Ftdd`\n  rules. Look at stages to make sure each of them keeps the tests green; in\n  case that would take too much temporary effort, allow red, note about test\n  status and compact reason for it.\n\n- Map out scope-specific limitations, follow-ups and related out-of-scope\n  concerns.\n\n- Prefer reuse and reduction of existing nouns, verbs and adjectives; when\n  adding shape - make sure documentation supports it, if in any doubt - `\u002Falign`\n  with user.\n\n- Any wrinkle against architecture should be resolved; when solution is not\n  found or following first-principles makes the code weird - align with user.\n  Architecture must lead code shape even if code disagrees. Do not expose\n  public methods or create flows that are not in architecture.\n\n- Each choice the RFC takes names the line of the governing question's accepted\n  `Where it stands` it keeps; a choice that keeps none, or crosses one, goes to\n  `\u002Falign` before it is planned.\n\n- If planned code includes a temporary solution that is dissolved by future\n  development, add RFC instruction to append TODO comment to code that flags\n  this. Otherwise code changes can start relying or considering the temporal\n  code, which can be hard to disentangle later.\n\n- Repeated invocations are validation passes over the same RFC. Re-read\n  the ticket, RFC, durable docs, and relevant code; challenge the plan against\n  new evidence and amend it in place. Do not create a replacement RFC merely\n  because the plan changed before implementation.\n\n- As an additional pass, try to explain things to yourself simply, as if you\n  are teaching the architecture, and being asked reasonable question and look\n  for areas that evade simple or common-sense explanation.\n\n- Validate the RFC adversarially on every pass:\n  - Probe each boundary and stage with counterexamples, including empty, partial,\n    faulting, and raced outcomes where relevant.\n  - Make every planned test prove the intended behavior and failure reason, not\n    merely that the path succeeds or raises.\n  - Separate sourced facts and existing invariants from assumptions; verify\n    facts in code or authoritative sources and surface load-bearing assumptions\n    for `\u002Falign`.\n  - Reject pseudocode or prescribed structure that introduces undocumented\n    architecture or freezes a reversible local choice.\n  - Check that every new promise names how it will be validated, and that\n    the stages collectively prove the ticket's acceptance criteria.\n\n- Repeat validation using this skill until no load-bearing ambiguity,\n  contradiction, or unverified assumption remains. The RFC may permit multiple\n  equivalent local implementations when they preserve the same documented shape\n  and proof.\n\n- If repeated validation keep bringing up gaps, take a step back and look\n  at what causes this oscillation. The architecture, ADRs, tickets are not\n  carved in stone, they can have real contradictions or inconsistencies that cause\n  it. Look into the core reasons and `\u002Falign` on them again if needed.\n\n- Overall, always consider future development and potential code reuse when\n  selecting code shapes.\n\n- Cover migrations, compatibility, rollout, failure handling, and observability\n  when relevant.\n\n- Record the plan as an RFC; its name and home are installed below.\n\n- RFC header: H1 is the ticket's title plus ` — implementation plan`; then\n  `- **record of** what it intends to become`, which the mover turns to\n  `what happened` at close; then `**Authored:** YYYY-MM-DD`, and `**Last amended:** YYYY-MM-DD` once a later\n  pass changes the plan; then one line stating what it implements, linking the\n  ticket.\n\n- DO NOT CHANGE THE CODE!\n\n## Installed from other mechanisms\n\n\u003Cinstalled by=\"ticket\">\n**P6** Name the RFC by the owning ticket's basename, under `docs\u002Frfc\u002F`, with no serial of its own; it\nmoves to `docs\u002Frfc\u002Fdone\u002F` with the ticket.\n\u003C\u002Finstalled>\n\n\u003Cinstalled by=\"session\">\n**S7** At the step's start, hold the unit by the question its work answers, `sessions.py --hold \u003Cq-id>\n--session \u003Ctag>`; a unit this session held is let go by it: say which, and that it is nobody's\nnow. A refusal naming another running session stops the step: say whose the unit is, and do\nnothing on it. A refusal saying the question is no unit leaves the step to go on, holding\nnothing.\n\u003C\u002Finstalled>\n",{"name":5,"description":7},"main","a43633f7ef0eeeb220f42872be487150daec8575",".agents\u002Fskills\u002Fplan","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness","https:\u002F\u002Fgithub.com\u002Fdveyarangi","http:\u002F\u002Fdveyarangi.blogspot.com\u002F","2026-11-14 12:09:52",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add dveyarangi\u002Fgoodwolf-harness --skill plan",{"tool":37,"command":38},"bunx","bunx skills add dveyarangi\u002Fgoodwolf-harness --skill plan",{"tool":40,"command":41},"pnpm","pnpm dlx skills add dveyarangi\u002Fgoodwolf-harness --skill plan","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fplan\u002FSKILL.md",true,[],[46,53,60,67,74,81],{"skillNo":47,"name":48,"nameZh":6,"description":49,"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},"01M7CDYHDFRAZP8T121PGMZ46R","verify","Use after an implementation pass, or to verify a landed slice before close. Whole verification of landed work: ticket, RFC, governing docs, and the project's verification set.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fverify\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M7CDYES9QW4N2JZ8GX1Q9DMV","ticket","Use to decompose a parent work into independently-workable child tickets. Use to mint ticket(s), possibly a single one, when planning implementation without one.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fticket\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M7CDYB6CFM7PD0APTBBS17HH","tdd","Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions \"red-green-refactor\", wants integration tests, or asks for test-first development.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftdd\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7CDY7V9VR5DDC0TPXQRB6K1","skill-up","Aid agent skill creation or modification. Use when creating a new skill or changing an existing one. Use whenever writing any agent instructions, whether to a file or directly in a reply.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fskill-up\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7CDY4B01RZP7Y90JGTCHNY5","setup-devops","Set up the project's engineering scaffolding, from toolchain to CICD and observability. Use when initializing a repo or filling a setup gap.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fsetup-devops\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M7CDY1P8T9VT2W2ZXEH1H4PR","review-architecture","Explore the codebase for module-deepening opportunities — architectural improvements that raise testability and navigability. Use when asked to improve or refactor the architecture.","https:\u002F\u002Fgithub.com\u002Fdveyarangi\u002Fgoodwolf-harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Freview-architecture\u002FSKILL.md",[],[]]