[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7B1MJ8EHWXNYWNX60WB00X5":3,"skill-same:01M7B1MJ8EHWXNYWNX60WB00X5":51},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":31,"install":32,"resources":44},"01M7B1MJ8EHWXNYWNX60WB00X5","writing-for-agents",null,"Reference for writing documents an agent consumes: skills, AGENTS.md or CLAUDE.md, and docs reached by a pointer. Invoke when creating or editing a skill or an agent instruction file, or when a document's pointer wording, load, or structure is in question. Do not invoke merely because the user mentions writing-for-agents, names a skill, or writes 'skill \u003Cword>'.","mineugene\u002Fpi-dev-config","mineugene","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F8313048?v=4","01M7B1HXRN5A0CS3ZH5MAR30VD","Eugene Min","01M7B1HXS9CJCHJK9KXDSYNRTK",1,"MIT","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-for-agents\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: writing-for-agents\ndisable-model-invocation: true\ndescription: \"Reference for writing documents an agent consumes: skills, AGENTS.md or CLAUDE.md, and docs reached by a pointer. Invoke when creating or editing a skill or an agent instruction file, or when a document's pointer wording, load, or structure is in question. Do not invoke merely because the user mentions writing-for-agents, names a skill, or writes 'skill \u003Cword>'.\"\n---\n# Writing for Agents\n\n## Contents\n\n- **Plain language** - the four ISO 24495-1 principles and how the skill's\n  levers reach them.\n- **Context pointers** - descriptions and links that trigger reference loading.\n- **The two loads** - agent context cost and human document recall.\n- **Information hierarchy** - steps, reference, disclosure, co-location, and\n  outlines.\n- **Steps and completion criteria** - checkable, exhaustive stopping points.\n- **When to split** - sequence and invocation boundaries.\n- **Leading words** - shared concepts and positive instructions.\n- **Pruning** - one source of truth, relevance, and removing no-ops.\n\nReference for writing any document an agent consumes: a skill, an `AGENTS.md` \u002F\n`CLAUDE.md`, or a doc reached by a pointer. The packaging differs; the writing\ndoes not: the same levers make each one predictable, since the agent takes the\nsame _process_ every run rather than producing the same output.\n\nWrite instructions and user-facing explanations in plain language. Keep precise\nterms, commands, and real limits; use code or mathematical notation only when\nthe subject needs it, not to frame ordinary workflows.\n\nWhen the document you're writing is a skill, read\n[SKILL-MECHANICS.md](.\u002FSKILL-MECHANICS.md) for frontmatter, invocation choice,\nand router skills.\n\n## Plain language\n\nPlain language is communication that puts the reader first. A document is plain\nwhen the reader can do four things on a first reading\n([ISO 24495-1:2023](https:\u002F\u002Fwww.iso.org\u002Fstandard\u002F78907.html)):\n\n- **Relevant**: get the information they need, and only that.\n- **Findable**: locate it without searching.\n- **Understandable**: understand it on the first pass.\n- **Usable**: act on it.\n\nTreat the four as the outcome to check, not a style to apply. The skill's own\nlevers already reach them: **relevant** is pruning, **findable** is context\npointers, **understandable** is the information hierarchy, and **usable** is\nsteps with checkable completion criteria.\n\n## Context pointers\n\nA **context pointer** is a reference held in the agent's context that names some\nout-of-context material and encodes the condition for reaching it. A skill's\ndescription is one; a line in `AGENTS.md` naming a doc is the same object. The\npointer's _wording_, not its target, decides when the agent reaches the\nmaterial, and how reliably. A must-have target behind a weakly worded pointer is\na variance bug: sharpen the wording first, and inline the material only if\nsharpening fails.\n\nA pointer does two jobs: state what the material is, and list the **branches**\nthat should trigger reaching it (a branch is a distinct case the document\nhandles, so different runs take different paths through it). Every word of an\nalways-loaded pointer costs on every turn, so it earns even harder pruning than\nthe body:\n\n- **Front-load the leading word**: the pointer is where it does its triggering\n  work.\n- **One trigger per branch.** Synonyms that rename a single branch are one\n  branch written twice; collapse them and keep only genuinely distinct branches.\n- **Cut identity the body already carries.**\n\n## The two loads\n\nEvery document and pointer you add spends one of two budgets:\n\n- **Context load** is the cost of always-loaded material on the agent's window:\n  an `AGENTS.md` line, a skill description, anything sitting in context every\n  turn, spending tokens and attention whether or not it fires.\n- **Cognitive load** is the cost on the human: which documents exist and when to\n  reach for each. The human is the index. Not a cost to minimise: it is the\n  price of human agency; spend it where human judgement matters, remove it where\n  it does not.\n\nMaterial reached only through a pointer escapes context load at the price of the\npointer's own line; material with no pointer at all rides entirely on cognitive\nload.\n\n## Information hierarchy\n\nA document is built from two content types: **steps** (the ordered actions the\nagent performs) and **reference** (definitions, rules, facts consulted on\ndemand). The two mix freely: all steps (a recipe), all reference (a review's\nrules, this skill), or both. The core decision is where each piece sits on the\n**information hierarchy**, a ladder ranked by how immediately the agent needs\nthe material:\n\n1. **In-file step** is the primary tier: what the agent does, in order.\n2. **In-file reference** is consulted on demand. Often a legitimately flat\n   peer-set (every rule of a review on one rung), which is a fine arrangement,\n   not a smell.\n3. **Disclosed reference** is pushed out into a separate file, reached by a\n   context pointer, loaded only when the pointer fires. Spans a sibling file in\n   the same folder through fully external reference that lives anywhere and any\n   document can point at.\n\nPush too little down and the top bloats; push too much and you hide material the\nagent actually needs. That tension is the whole decision.\n\n**Progressive disclosure** is the move down the ladder (out of the main file and\nbehind a pointer) so the top stays legible. Not primarily a token optimisation:\nit is how the hierarchy is protected. Branching is the cleanest disclosure test:\ninline what every branch needs, and push behind a pointer what only some\nbranches reach. When a document has steps, in-file reference that should be\ndisclosed buries them and turns attending to them into a coin-flip: a variance\nlever, not just a legibility one.\n\n**Co-location** is the within-file companion: where the ladder decides _how far\ndown_ a piece sits, co-location decides _what sits beside it_ once there. Keep a\nconcept's definition, rules, and caveats under one heading rather than\nscattered, so reading one part brings its neighbours with it. The test: the\ndocument should read like documentation written for the agent. Grouped material\nreads that way; scattered material does not. (Distinct from duplication: that\nrepeats one meaning in two places; scattering fragments one meaning across\nmany.)\n\n**Sprawl** is the failure mode here: a document simply too long, even when every\nline is live and unique. Attention thins across the excess, and every extra line\nis one more to keep relevant. The cure is the ladder: disclose reference behind\npointers, and split by branch or sequence so each path carries only what it\nneeds.\n\n**Outline**: a document over 200 lines must open with a `## Contents` section.\nIt is a bulleted list in document order, one item per top-level section,\nformatted `- **\u003Cheading>** - \u003Cthe section's keywords and one-line description>`,\nso a read truncated at 200 lines still reveals what follows.\n\n## Steps and completion criteria\n\nEvery step ends on a **completion criterion**, the condition that tells the\nagent the work is done. Two properties make it a lever:\n\n- **Clarity**: can the agent tell done from not-done? A vague bound\n  (\"understanding reached\") invites **premature completion**: ending the step\n  before it is genuinely done, attention slipping to _being done_. The visible\n  steps still ahead (the **post-completion steps**) supply the pull; the\n  criterion's clarity is the resistance. Defend in order: **sharpen the bound\n  first** (local and cheap); only if it is irreducibly fuzzy _and_ you observe\n  the rush, hide the later steps by splitting the sequence. Hiding only works\n  across a real context boundary (a hand-off or a subagent dispatch; an inline\n  call leaves the later steps in context and clears nothing).\n- **Demand**: how much it requires. \"Every modified model accounted for\" forces\n  thorough work where \"produce a change list\" does not. Demand drives\n  **legwork** (the digging the agent does within the work, latent in the wording\n  rather than written as its own step), and it is not step-bound: \"every rule\n  applied\" binds a body of flat reference just as \"every step done\" binds a\n  sequence, which is how an all-reference document still carries an\n  exhaustiveness bar.\n\nThe strongest criteria are both checkable and exhaustive.\n\n## When to split\n\nSplitting one document into two spends one of the two loads, so split only when\nthe cut earns it:\n\n- **By sequence**: split a run of steps where the post-completion steps tempt\n  the agent to rush the one in front of it. Keeping them out of view drives more\n  legwork on the current task. Beware the reverse: merging sequences exposes\n  each step's later steps to what follows, inviting premature completion.\n- **By invocation**, skill-specific: see\n  [SKILL-MECHANICS.md](.\u002FSKILL-MECHANICS.md).\n\n## Leading words\n\nA **leading word** is a compact concept already living in the model's\npretraining that the agent thinks with while running the document (_lesson_,\n_fog of war_, _tracer bullets_). Repeated as a token, never as a sentence, it\naccumulates a distributed definition and anchors a whole region of behaviour in\nthe fewest tokens, by recruiting priors the model already holds. Coining your\nown works if you define it clearly, but a made-up word recruits no priors: you\npay in definition tokens what a pretrained word gives free; reach for an\nexisting word first.\n\nIt anchors twice. In the body, _execution_: the agent reaches for the same\nbehaviour every time the word appears, and inside flat reference it focuses\nattention on a class of thing to look for. In a pointer, _invocation_: when the\nsame word lives in your prompts, your docs, and your codebase, the agent links\nthat shared language to the material and reaches it more reliably.\n\nHunt for opportunities to refactor with leading words. A triad spelled out at\nthree sites, a pointer spending a sentence to gesture at one idea. Each is a\npassage begging to collapse into a single token:\n\n- \"fast, deterministic, low-overhead\" becomes _tight_ (a _tight_ loop).\n- \"a loop you believe in\" becomes _red_, turning a fuzzy gate into a binary\n  observable state (the loop goes _red_ on the bug, or it doesn't).\n\nYou win twice: fewer tokens, and a sharper hook for the agent to hang its\nthinking on. Assume every document is carrying restatements that leading words\nretire. Go find them.\n\n**Negation** is the failure mode beside this lever: steering by prohibition\ndrags the forbidden behaviour into context and makes it _more_ available, not\nless. _Don't think of an elephant_, and the elephant is all there is; the\nnegation is a weak modifier the strongly-activated concept overruns, so the ban\nhalf-reads as an instruction to do the thing. Prompt the **positive**: state the\ntarget behaviour (\"write one-line comments\") so the banned one is never spoken.\nA prohibition earns its place only as a hard guardrail you cannot phrase\npositively; even then, pair it with the positive target so attention lands on\nwhat to do.\n\n## Pruning\n\n- Keep each meaning in a **single source of truth**: one authoritative place, so\n  changing the behaviour is a one-place edit. **Duplication** (the same meaning\n  in more than one place) costs maintenance and tokens, and inflates a meaning's\n  prominence on the ladder past its real rank. (The accidental inverse of a\n  leading word, which repeats a token on purpose, never the meaning.)\n- The **environment** is a source of truth too (`package.json` scripts, config\n  files, the directory layout, `--help` output), and a document that restates it\n  is a **cache**: a copy of a lookup, earning its load only when the lookup is\n  expensive. Cache what the agent cannot find by looking: the unwritten\n  convention, the reason behind a choice, the gotcha no config confesses. Leave\n  the one-file, one-command lookups to the environment, where they cannot go\n  stale.\n- Check every line for **relevance**: does it still bear on what the document\n  does? A line loses relevance by never bearing on the task (mere exposition, or\n  a branch that should be disclosed) or by going stale as the behaviour or world\n  it describes changes. Shorter documents are easier to keep relevant. Without a\n  pruning discipline the default fate is **sediment**: stale layers that settle\n  because adding feels safe and removing feels risky, until you must core down\n  through them to find what is still live.\n- Hunt **no-ops** sentence by sentence: an instruction the model already obeys\n  by default pays load to say nothing. The test (does it change behaviour versus\n  the default?) is model-relative, not reader-relative: two people disagreeing\n  about a no-op disagree about the default, and settle it by running the\n  document, not by debate. When a sentence fails, delete the whole sentence\n  rather than trim words from it. The test also grades leading words: a word too\n  weak to beat the default (_be thorough_ when the agent is already\n  thorough-ish) is a no-op, and the fix is a stronger word (_relentless_), not a\n  different technique.\n","# Writing for Agents\n\n## Contents\n\n- **Plain language** - the four ISO 24495-1 principles and how the skill's\n  levers reach them.\n- **Context pointers** - descriptions and links that trigger reference loading.\n- **The two loads** - agent context cost and human document recall.\n- **Information hierarchy** - steps, reference, disclosure, co-location, and\n  outlines.\n- **Steps and completion criteria** - checkable, exhaustive stopping points.\n- **When to split** - sequence and invocation boundaries.\n- **Leading words** - shared concepts and positive instructions.\n- **Pruning** - one source of truth, relevance, and removing no-ops.\n\nReference for writing any document an agent consumes: a skill, an `AGENTS.md` \u002F\n`CLAUDE.md`, or a doc reached by a pointer. The packaging differs; the writing\ndoes not: the same levers make each one predictable, since the agent takes the\nsame _process_ every run rather than producing the same output.\n\nWrite instructions and user-facing explanations in plain language. Keep precise\nterms, commands, and real limits; use code or mathematical notation only when\nthe subject needs it, not to frame ordinary workflows.\n\nWhen the document you're writing is a skill, read\n[SKILL-MECHANICS.md](.\u002FSKILL-MECHANICS.md) for frontmatter, invocation choice,\nand router skills.\n\n## Plain language\n\nPlain language is communication that puts the reader first. A document is plain\nwhen the reader can do four things on a first reading\n([ISO 24495-1:2023](https:\u002F\u002Fwww.iso.org\u002Fstandard\u002F78907.html)):\n\n- **Relevant**: get the information they need, and only that.\n- **Findable**: locate it without searching.\n- **Understandable**: understand it on the first pass.\n- **Usable**: act on it.\n\nTreat the four as the outcome to check, not a style to apply. The skill's own\nlevers already reach them: **relevant** is pruning, **findable** is context\npointers, **understandable** is the information hierarchy, and **usable** is\nsteps with checkable completion criteria.\n\n## Context pointers\n\nA **context pointer** is a reference held in the agent's context that names some\nout-of-context material and encodes the condition for reaching it. A skill's\ndescription is one; a line in `AGENTS.md` naming a doc is the same object. The\npointer's _wording_, not its target, decides when the agent reaches the\nmaterial, and how reliably. A must-have target behind a weakly worded pointer is\na variance bug: sharpen the wording first, and inline the material only if\nsharpening fails.\n\nA pointer does two jobs: state what the material is, and list the **branches**\nthat should trigger reaching it (a branch is a distinct case the document\nhandles, so different runs take different paths through it). Every word of an\nalways-loaded pointer costs on every turn, so it earns even harder pruning than\nthe body:\n\n- **Front-load the leading word**: the pointer is where it does its triggering\n  work.\n- **One trigger per branch.** Synonyms that rename a single branch are one\n  branch written twice; collapse them and keep only genuinely distinct branches.\n- **Cut identity the body already carries.**\n\n## The two loads\n\nEvery document and pointer you add spends one of two budgets:\n\n- **Context load** is the cost of always-loaded material on the agent's window:\n  an `AGENTS.md` line, a skill description, anything sitting in context every\n  turn, spending tokens and attention whether or not it fires.\n- **Cognitive load** is the cost on the human: which documents exist and when to\n  reach for each. The human is the index. Not a cost to minimise: it is the\n  price of human agency; spend it where human judgement matters, remove it where\n  it does not.\n\nMaterial reached only through a pointer escapes context load at the price of the\npointer's own line; material with no pointer at all rides entirely on cognitive\nload.\n\n## Information hierarchy\n\nA document is built from two content types: **steps** (the ordered actions the\nagent performs) and **reference** (definitions, rules, facts consulted on\ndemand). The two mix freely: all steps (a recipe), all reference (a review's\nrules, this skill), or both. The core decision is where each piece sits on the\n**information hierarchy**, a ladder ranked by how immediately the agent needs\nthe material:\n\n1. **In-file step** is the primary tier: what the agent does, in order.\n2. **In-file reference** is consulted on demand. Often a legitimately flat\n   peer-set (every rule of a review on one rung), which is a fine arrangement,\n   not a smell.\n3. **Disclosed reference** is pushed out into a separate file, reached by a\n   context pointer, loaded only when the pointer fires. Spans a sibling file in\n   the same folder through fully external reference that lives anywhere and any\n   document can point at.\n\nPush too little down and the top bloats; push too much and you hide material the\nagent actually needs. That tension is the whole decision.\n\n**Progressive disclosure** is the move down the ladder (out of the main file and\nbehind a pointer) so the top stays legible. Not primarily a token optimisation:\nit is how the hierarchy is protected. Branching is the cleanest disclosure test:\ninline what every branch needs, and push behind a pointer what only some\nbranches reach. When a document has steps, in-file reference that should be\ndisclosed buries them and turns attending to them into a coin-flip: a variance\nlever, not just a legibility one.\n\n**Co-location** is the within-file companion: where the ladder decides _how far\ndown_ a piece sits, co-location decides _what sits beside it_ once there. Keep a\nconcept's definition, rules, and caveats under one heading rather than\nscattered, so reading one part brings its neighbours with it. The test: the\ndocument should read like documentation written for the agent. Grouped material\nreads that way; scattered material does not. (Distinct from duplication: that\nrepeats one meaning in two places; scattering fragments one meaning across\nmany.)\n\n**Sprawl** is the failure mode here: a document simply too long, even when every\nline is live and unique. Attention thins across the excess, and every extra line\nis one more to keep relevant. The cure is the ladder: disclose reference behind\npointers, and split by branch or sequence so each path carries only what it\nneeds.\n\n**Outline**: a document over 200 lines must open with a `## Contents` section.\nIt is a bulleted list in document order, one item per top-level section,\nformatted `- **\u003Cheading>** - \u003Cthe section's keywords and one-line description>`,\nso a read truncated at 200 lines still reveals what follows.\n\n## Steps and completion criteria\n\nEvery step ends on a **completion criterion**, the condition that tells the\nagent the work is done. Two properties make it a lever:\n\n- **Clarity**: can the agent tell done from not-done? A vague bound\n  (\"understanding reached\") invites **premature completion**: ending the step\n  before it is genuinely done, attention slipping to _being done_. The visible\n  steps still ahead (the **post-completion steps**) supply the pull; the\n  criterion's clarity is the resistance. Defend in order: **sharpen the bound\n  first** (local and cheap); only if it is irreducibly fuzzy _and_ you observe\n  the rush, hide the later steps by splitting the sequence. Hiding only works\n  across a real context boundary (a hand-off or a subagent dispatch; an inline\n  call leaves the later steps in context and clears nothing).\n- **Demand**: how much it requires. \"Every modified model accounted for\" forces\n  thorough work where \"produce a change list\" does not. Demand drives\n  **legwork** (the digging the agent does within the work, latent in the wording\n  rather than written as its own step), and it is not step-bound: \"every rule\n  applied\" binds a body of flat reference just as \"every step done\" binds a\n  sequence, which is how an all-reference document still carries an\n  exhaustiveness bar.\n\nThe strongest criteria are both checkable and exhaustive.\n\n## When to split\n\nSplitting one document into two spends one of the two loads, so split only when\nthe cut earns it:\n\n- **By sequence**: split a run of steps where the post-completion steps tempt\n  the agent to rush the one in front of it. Keeping them out of view drives more\n  legwork on the current task. Beware the reverse: merging sequences exposes\n  each step's later steps to what follows, inviting premature completion.\n- **By invocation**, skill-specific: see\n  [SKILL-MECHANICS.md](.\u002FSKILL-MECHANICS.md).\n\n## Leading words\n\nA **leading word** is a compact concept already living in the model's\npretraining that the agent thinks with while running the document (_lesson_,\n_fog of war_, _tracer bullets_). Repeated as a token, never as a sentence, it\naccumulates a distributed definition and anchors a whole region of behaviour in\nthe fewest tokens, by recruiting priors the model already holds. Coining your\nown works if you define it clearly, but a made-up word recruits no priors: you\npay in definition tokens what a pretrained word gives free; reach for an\nexisting word first.\n\nIt anchors twice. In the body, _execution_: the agent reaches for the same\nbehaviour every time the word appears, and inside flat reference it focuses\nattention on a class of thing to look for. In a pointer, _invocation_: when the\nsame word lives in your prompts, your docs, and your codebase, the agent links\nthat shared language to the material and reaches it more reliably.\n\nHunt for opportunities to refactor with leading words. A triad spelled out at\nthree sites, a pointer spending a sentence to gesture at one idea. Each is a\npassage begging to collapse into a single token:\n\n- \"fast, deterministic, low-overhead\" becomes _tight_ (a _tight_ loop).\n- \"a loop you believe in\" becomes _red_, turning a fuzzy gate into a binary\n  observable state (the loop goes _red_ on the bug, or it doesn't).\n\nYou win twice: fewer tokens, and a sharper hook for the agent to hang its\nthinking on. Assume every document is carrying restatements that leading words\nretire. Go find them.\n\n**Negation** is the failure mode beside this lever: steering by prohibition\ndrags the forbidden behaviour into context and makes it _more_ available, not\nless. _Don't think of an elephant_, and the elephant is all there is; the\nnegation is a weak modifier the strongly-activated concept overruns, so the ban\nhalf-reads as an instruction to do the thing. Prompt the **positive**: state the\ntarget behaviour (\"write one-line comments\") so the banned one is never spoken.\nA prohibition earns its place only as a hard guardrail you cannot phrase\npositively; even then, pair it with the positive target so attention lands on\nwhat to do.\n\n## Pruning\n\n- Keep each meaning in a **single source of truth**: one authoritative place, so\n  changing the behaviour is a one-place edit. **Duplication** (the same meaning\n  in more than one place) costs maintenance and tokens, and inflates a meaning's\n  prominence on the ladder past its real rank. (The accidental inverse of a\n  leading word, which repeats a token on purpose, never the meaning.)\n- The **environment** is a source of truth too (`package.json` scripts, config\n  files, the directory layout, `--help` output), and a document that restates it\n  is a **cache**: a copy of a lookup, earning its load only when the lookup is\n  expensive. Cache what the agent cannot find by looking: the unwritten\n  convention, the reason behind a choice, the gotcha no config confesses. Leave\n  the one-file, one-command lookups to the environment, where they cannot go\n  stale.\n- Check every line for **relevance**: does it still bear on what the document\n  does? A line loses relevance by never bearing on the task (mere exposition, or\n  a branch that should be disclosed) or by going stale as the behaviour or world\n  it describes changes. Shorter documents are easier to keep relevant. Without a\n  pruning discipline the default fate is **sediment**: stale layers that settle\n  because adding feels safe and removing feels risky, until you must core down\n  through them to find what is still live.\n- Hunt **no-ops** sentence by sentence: an instruction the model already obeys\n  by default pays load to say nothing. The test (does it change behaviour versus\n  the default?) is model-relative, not reader-relative: two people disagreeing\n  about a no-op disagree about the default, and settle it by running the\n  document, not by debate. When a sentence fails, delete the whole sentence\n  rather than trim words from it. The test also grades leading words: a word too\n  weak to beat the default (_be thorough_ when the agent is already\n  thorough-ish) is a no-op, and the fix is a stronger word (_relentless_), not a\n  different technique.\n",{"name":5,"description":7,"disable-model-invocation":24},true,"main","656f9af15679b3c6a7df060a321bf853acb56e8c","skills\u002Fwriting-for-agents",0,"https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config","https:\u002F\u002Fgithub.com\u002Fmineugene","2026-11-12 07:59:18",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":24},[34,37,40],{"tool":35,"command":36},"npx","npx skills add mineugene\u002Fpi-dev-config --skill writing-for-agents",{"tool":38,"command":39},"bunx","bunx skills add mineugene\u002Fpi-dev-config --skill writing-for-agents",{"tool":41,"command":42},"pnpm","pnpm dlx skills add mineugene\u002Fpi-dev-config --skill writing-for-agents","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-for-agents\u002FSKILL.md",[45],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"SKILL-MECHANICS.md","REFERENCE","参考文档",2657,"https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-for-agents\u002FSKILL-MECHANICS.md",[52,59,66,73,80,87],{"skillNo":53,"name":54,"nameZh":6,"description":55,"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},"01M7B1MDTK7MDRPNFPRM7ATT23","wizard","Generate an interactive bash wizard for steps only a human can perform. Invoke when the user asks for a script or wizard for manual setup: infrastructure provisioning, credentials or CI secrets, a third-party dashboard, or a one-off migration or cutover. Do not invoke for steps the agent can perform, or merely because the user mentions wizard, names a skill, or writes 'skill \u003Cword>'.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fwizard\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M7B1MAHMQPAEE4Q8AZF8WW9N","wait-what","Stop and re-pitch the last message in plain, simplified language.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fwait-what\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M7B1M8NS8GYP1QDEYY9KXPQP","to-tickets","Split an agreed specification into reviewable vertical-slice issues with explicit blockers, then publish them after confirmation.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fto-tickets\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7B1M6M89D3KH0QTB71061QX","to-spec","Turn the current conversation into a specification and publish it after confirmation.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fto-spec\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7B1M540CQGT30DHPKZ2PCQF","to-pr","Create a pull request from the current branch.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Fto-pr\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":19},"01M7B1M2X12G3FBER7DP21DRDX","tdd","Test-driven development. Invoke only when the user asks to build a feature or fix a bug test-first, requests a red-green-refactor loop, or requests integration tests.","https:\u002F\u002Fgithub.com\u002Fmineugene\u002Fpi-dev-config\u002Fblob\u002Fmain\u002Fskills\u002Ftdd\u002FSKILL.md",[],[]]