@warlock.js/ai 4.6.0 → 4.7.0

package/esm/system-prompt/system-prompt.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"system-prompt.mjs","names":[],"sources":["../../../../../../../@warlock.js/ai/src/system-prompt/system-prompt.ts"],"sourcesContent":["import { readFileSync } from \"node:fs\";\nimport type { Placeholders } from \"../contracts/placeholders.type\";\nimport type {\n  InstructionContract,\n  PersonaContract,\n  SystemPromptBlockContract,\n  SystemPromptContract,\n  SystemPromptMergeOptions,\n  SystemPromptMeta,\n} from \"../contracts/system-prompt.contract\";\nimport { InvalidRequestError } from \"../errors\";\nimport { defaultPromptsManager, promptKey } from \"../prompts/prompts-manager\";\nimport type {\n  PromptValidationResult,\n  PromptsValidateOptions,\n} from \"../prompts/prompts-manager.type\";\nimport { Instruction } from \"./instruction\";\nimport { Persona } from \"./persona\";\n\n/**\n * Monotonic source of the internal, non-registry display id every\n * `SystemPrompt` carries. Anonymous (unnamed) prompts have nothing else to\n * identify them by; this id never feeds the registry and is never derived from\n * the wall clock, so it stays stable and order-deterministic across a run.\n */\nlet displayIdCounter = 0;\n\n/**\n * Narrow an arbitrary value to a `SystemPromptContract` — true when it exposes\n * the builder surface (`blocks` array + a callable `resolve`). Used by the\n * registry-aware `merge` overload to tell a folded contract from a raw block\n * or a registry name string, robustly across duplicate package copies.\n */\nfunction isSystemPromptContract(\n  value: unknown,\n): value is SystemPromptContract {\n  return (\n    typeof value === \"object\" &&\n    value !== null &&\n    Array.isArray((value as { blocks?: unknown }).blocks) &&\n    typeof (value as { resolve?: unknown }).resolve === \"function\"\n  );\n}\n\n/**\n * Build the deterministic provenance label for a prompt — `name@version` when\n * it is registered, otherwise its internal display id. No random suffixes, so\n * the same source always yields the same `composedFrom` entry.\n */\nfunction provenanceLabel(prompt: SystemPromptContract): string {\n  const meta = prompt.meta();\n\n  if (meta?.name) {\n    return promptKey(meta.name, meta.version ?? \"1\");\n  }\n\n  return prompt instanceof SystemPrompt ? prompt.id : \"anonymous\";\n}\n\n/**\n * Concrete `SystemPromptContract` — an immutable layered prompt builder.\n *\n * **Role.** The top-level composer for a system prompt: it holds an ordered\n * list of typed blocks (persona + instructions) and resolves the whole\n * stack into one final string when the agent is about to call the model.\n *\n * **Responsibility.**\n * - Owns: the ordered `blocks` list and the block-join rules (insertion\n *   order, blank-line separator, trim).\n * - Does NOT own: how any individual block is rendered (delegated to each\n *   block's `resolve()`), the placeholder syntax (delegated to\n *   `renderPlaceholders`), or any knowledge of the agent, model, or\n *   session consuming the resolved text.\n *\n * Blocks are discriminated by a string `type` tag (`\"persona\"` /\n * `\"instruction\"`) rather than `instanceof`, so user-supplied blocks that\n * implement `SystemPromptBlockContract` interoperate seamlessly with blocks\n * built via `ai.persona()` / `ai.instruction()` — even across duplicate\n * package copies or bundler scope boundaries.\n *\n * The builder is **immutable** — every `.persona()` / `.instruction()`\n * call returns a fresh `SystemPrompt` instance sharing nothing mutable\n * with its parent. This makes forking a base prompt into specialized\n * variants a safe, side-effect-free operation.\n *\n * Users construct via the `ai.systemPrompt()` factory — `new SystemPrompt()`\n * is not the public API (see §4.2 of code-style.md). Modeled as a class so\n * that methods live on the prototype (one copy shared across every forked\n * instance) and downstream code can branch via `instanceof SystemPrompt`.\n *\n * @example\n * // Chainable form\n * const alex = ai.persona(\"You are Alex, a TypeScript expert.\");\n * const replyIn = ai.instruction(\"Respond in {{language|English}}.\");\n *\n * const base = ai.systemPrompt().persona(alex).instruction(replyIn);\n * const arabicVariant = base.instruction(\"Prefer Arabic comments.\");\n *\n * base.resolve({ language: \"English\" });\n * arabicVariant.resolve({ language: \"Arabic\" });\n *\n * @example\n * // Array form — insertion order is preserved exactly\n * const prompt = ai.systemPrompt([\n *   ai.persona(\"You are Alex, a TypeScript expert.\"),\n *   ai.instruction(\"Respond in {{language|English}}.\"),\n * ]);\n */\nexport class SystemPrompt implements SystemPromptContract {\n  /**\n   * Internal, non-registry id for display / provenance. Stable for the life of\n   * the instance; sourced from a monotonic counter, never the wall clock.\n   * Anonymous prompts are identified solely by this id.\n   */\n  public readonly id: string;\n\n  public constructor(\n    public readonly blocks: readonly SystemPromptBlockContract[] = [],\n    private readonly metaData?: SystemPromptMeta,\n  ) {\n    this.id = `prompt#${displayIdCounter++}`;\n\n    // Auto-register the moment a builder acquires a name — whether through the\n    // `systemPrompt(input, { name })` factory or a `.meta({ name })` rename.\n    // Forks built by `persona()` / `instruction()` / `merge()` deliberately\n    // drop the name (they pass no meta), so they stay anonymous and never land\n    // in the registry unless explicitly re-named.\n    if (metaData?.name) {\n      defaultPromptsManager().register(this);\n    }\n  }\n\n  /**\n   * Read the current metadata snapshot (no argument) or derive a renamed\n   * builder (with `meta`). The accessor returns `undefined` for an anonymous\n   * prompt; the updater shallow-merges `meta` onto the current metadata and\n   * returns a fresh builder. Naming the result registers it in `ai.prompts`.\n   */\n  public meta(): SystemPromptMeta | undefined;\n  public meta(meta: SystemPromptMeta): SystemPromptContract;\n  public meta(\n    meta?: SystemPromptMeta,\n  ): SystemPromptMeta | undefined | SystemPromptContract {\n    if (meta === undefined) {\n      return this.metaData;\n    }\n\n    return new SystemPrompt(this.blocks, { ...this.metaData, ...meta });\n  }\n\n  /**\n   * Build a system prompt by reading the file at `path` once, synchronously,\n   * at construction time. The file's UTF-8 contents seed a single instruction\n   * block — the same semantics as the string-seed form of `systemPrompt()` —\n   * so placeholders inside the file (`{{language|English}}`) resolve at\n   * `resolve()` time and the result can be forked with further\n   * `.persona()` / `.instruction()` calls.\n   *\n   * One-shot by design: the file is read exactly once here, never re-read on\n   * `resolve()`. Reads are synchronous so the call stays a drop-in for the\n   * synchronous `systemPrompt()` factory and the synchronous `resolve()` API.\n   *\n   * Throws `InvalidRequestError` when the file cannot be read (missing path,\n   * permission denied) — surfacing the underlying cause so a typo in the\n   * prompt path fails loudly at construction instead of silently producing an\n   * empty prompt.\n   *\n   * @param path - Filesystem path to the prompt template file.\n   *\n   * @example\n   * const prompt = SystemPrompt.fromFile(\"./prompts/support-agent.md\");\n   *\n   * const localized = prompt.instruction(\"Respond in {{language|English}}.\");\n   * localized.resolve({ language: \"Arabic\" });\n   */\n  public static fromFile(path: string): SystemPrompt {\n    let contents: string;\n\n    try {\n      contents = readFileSync(path, \"utf8\");\n    } catch (error) {\n      throw new InvalidRequestError(\n        `Failed to read system prompt file \"${path}\" — ${\n          error instanceof Error ? error.message : String(error)\n        }`,\n        { context: { path }, cause: error },\n      );\n    }\n\n    return new SystemPrompt([new Instruction(contents)]);\n  }\n\n  /**\n   * Return a new builder with the persona block set. If a persona already\n   * exists it's replaced in place (preserving its position in `blocks`);\n   * otherwise the new persona is prepended so persona-first remains the\n   * default for chain-built prompts. Accepts either raw text (auto-wrapped\n   * via `new Persona`) or an existing `PersonaContract` instance for reuse\n   * across prompts.\n   */\n  public persona(value: PersonaContract | string): SystemPromptContract {\n    const block = typeof value === \"string\" ? new Persona(value) : value;\n    const existingIndex = this.blocks.findIndex(\n      candidate => candidate.type === \"persona\",\n    );\n\n    if (existingIndex >= 0) {\n      const next = [...this.blocks];\n      next[existingIndex] = block;\n\n      return new SystemPrompt(next) as this;\n    }\n\n    return new SystemPrompt([block, ...this.blocks]);\n  }\n\n  /**\n   * Return a new builder with the given instruction appended. Instructions\n   * render in insertion order. Accepts either raw text (auto-wrapped via\n   * `new Instruction`) or an existing `InstructionContract` instance for\n   * cross-prompt reuse.\n   */\n  public instruction(\n    value: InstructionContract | string,\n  ): SystemPromptContract {\n    const block = typeof value === \"string\" ? new Instruction(value) : value;\n\n    return new SystemPrompt([...this.blocks, block]);\n  }\n\n  /**\n   * Fold predefined blocks, another prompt contract, or a registered prompt\n   * name into this builder. Three forms share one method:\n   *\n   * - `merge(...blocks)` — N pre-built `ai.persona()` / `ai.instruction()`\n   *   blocks. A `persona` block sets/replaces the single, leading persona;\n   *   every other block appends in order. `base.merge(reviewer, style, lang)`\n   *   equals `base.persona(reviewer).instruction(style).instruction(lang)`.\n   * - `merge(contract)` — another prompt; its blocks fold in (persona\n   *   replaces, instructions append) and `meta.composedFrom` records the\n   *   provenance of both sides.\n   * - `merge(name, { fromVersion })` — a prompt resolved from `ai.prompts`\n   *   (latest version unless `fromVersion` selects another); throws\n   *   `InvalidRequestError` when the name / version is unregistered.\n   *\n   * Immutable — the original builder is untouched; passing zero blocks returns\n   * an equivalent builder. The folded result is anonymous (no `name`), so it\n   * is never auto-registered even though it carries `composedFrom` provenance.\n   */\n  public merge(\n    ...blocks: readonly SystemPromptBlockContract[]\n  ): SystemPromptContract;\n  public merge(source: SystemPromptContract): SystemPromptContract;\n  public merge(\n    name: string,\n    options?: SystemPromptMergeOptions,\n  ): SystemPromptContract;\n  public merge(\n    first?:\n      | SystemPromptBlockContract\n      | SystemPromptContract\n      | string,\n    // `undefined` is part of the element union so the `merge(name, options?)`\n    // overload's optional trailing `options?` (i.e. `… | undefined`) stays\n    // assignable to this implementation signature.\n    ...rest: readonly (\n      | SystemPromptBlockContract\n      | SystemPromptMergeOptions\n      | undefined\n    )[]\n  ): SystemPromptContract {\n    // Registry-name form: resolve from ai.prompts at the chosen version.\n    if (typeof first === \"string\") {\n      const options = rest[0] as SystemPromptMergeOptions | undefined;\n      const resolved = defaultPromptsManager().get(first, options?.fromVersion);\n\n      return this.mergeContract(resolved);\n    }\n\n    // Contract form: fold another prompt's blocks + record provenance.\n    if (isSystemPromptContract(first)) {\n      return this.mergeContract(first);\n    }\n\n    // Variadic block form (the original behavior).\n    const all = [\n      ...(first ? [first] : []),\n      ...rest,\n    ] as readonly SystemPromptBlockContract[];\n\n    return this.foldBlocks(this, all);\n  }\n\n  /**\n   * Fold an ordered list of blocks onto a starting prompt: persona blocks\n   * set/replace the single leading persona; every other block appends in\n   * order. The shared core of the variadic-block `merge` and the contract fold.\n   */\n  private foldBlocks(\n    start: SystemPromptContract,\n    blocks: readonly SystemPromptBlockContract[],\n  ): SystemPromptContract {\n    return blocks.reduce<SystemPromptContract>((prompt, block) => {\n      if (block.type === \"persona\") {\n        return prompt.persona(block as PersonaContract);\n      }\n\n      return new SystemPrompt([...prompt.blocks, block]);\n    }, start);\n  }\n\n  /**\n   * Fold another prompt contract into this one (persona replaces, instructions\n   * append) and stamp the deterministic `composedFrom` provenance — this\n   * prompt's existing provenance (or its own label) followed by the folded\n   * source's label. The result is anonymous so it never auto-registers.\n   */\n  private mergeContract(\n    source: SystemPromptContract,\n  ): SystemPromptContract {\n    const folded = this.foldBlocks(this, source.blocks);\n\n    const baseProvenance =\n      this.metaData?.composedFrom ??\n      (this.metaData?.name ? [provenanceLabel(this)] : []);\n\n    const composedFrom = [...baseProvenance, provenanceLabel(source)];\n\n    // Carry forward only provenance — never the name — so the merged result is\n    // a fresh anonymous prompt (immutable rename = new key; original stays).\n    return new SystemPrompt(folded.blocks, { composedFrom });\n  }\n\n  /**\n   * Resolve every block against the placeholder map, join the results with\n   * blank-line separators (in insertion order), and trim. Returns an empty\n   * string when no blocks are present — callers treat that as \"no system\n   * message\".\n   */\n  public resolve(placeholders?: Placeholders): string {\n    return this.blocks\n      .map(block => block.resolve(placeholders))\n      .join(\"\\n\\n\")\n      .trim();\n  }\n\n  /**\n   * Validate this prompt via the process-wide `ai.prompts` manager — sugar for\n   * `ai.prompts.validate(this, options)`. Runs the deterministic placeholder\n   * check and, when `options.judge` is supplied, the Nova-safe LLM-as-judge\n   * pass. Never throws on a judge failure; `ok` tracks the deterministic\n   * verdict alone.\n   */\n  public validate(\n    options?: PromptsValidateOptions,\n  ): Promise<PromptValidationResult> {\n    return defaultPromptsManager().validate(this, options);\n  }\n}\n\n/**\n * Public factory for `SystemPrompt`, callable directly or via its\n * `fromFile` static. Exists as a named interface so the callable signature\n * and the `fromFile` attachment travel together as one public type.\n */\nexport interface SystemPromptFactory {\n  (\n    input?: string | ReadonlyArray<SystemPromptBlockContract>,\n    meta?: SystemPromptMeta,\n  ): SystemPrompt;\n\n  /**\n   * Build a system prompt from a file read once at construction. Delegates\n   * to {@link SystemPrompt.fromFile}, so `ai.systemPrompt.fromFile(path)` and\n   * `SystemPrompt.fromFile(path)` behave identically.\n   *\n   * @example\n   * const prompt = ai.systemPrompt.fromFile(\"./prompts/support-agent.md\");\n   */\n  fromFile(path: string): SystemPrompt;\n}\n\nfunction systemPromptFactory(\n  input?: string | ReadonlyArray<SystemPromptBlockContract>,\n  meta?: SystemPromptMeta,\n): SystemPrompt {\n  if (input === undefined) {\n    return new SystemPrompt([], meta);\n  }\n\n  if (typeof input === \"string\") {\n    return new SystemPrompt([new Instruction(input)], meta);\n  }\n\n  return new SystemPrompt([...input], meta);\n}\n\n/**\n * Create a new immutable system-prompt builder.\n *\n * **Role.** Public factory for `SystemPrompt` — keeps user-facing code\n * free of `new` and consistent with `ai.tool()`, `ai.agent()`,\n * `ai.persona()`, `ai.instruction()`.\n *\n * Input forms:\n * - No argument → empty builder, chain `.persona()` / `.instruction()`\n * - Single string → seeded with one instruction for quick one-shot prompts\n * - Array of blocks → used verbatim, preserving insertion order\n * - `.fromFile(path)` → seeded from a file read once at construction\n *\n * Pass a second `meta` argument to name the prompt — a named prompt\n * auto-registers in `ai.prompts` under `name@version` (version defaults to the\n * next integer). Forks (`.persona()`, `.instruction()`, `.merge()`) are\n * anonymous unless re-named via `.meta({ name })`.\n *\n * @example\n * // Composed builder\n * const prompt = systemPrompt()\n *   .persona(\"You are Alex, a senior TypeScript engineer.\")\n *   .instruction(\"Always include working code examples.\")\n *   .instruction(\"Respond in {{language|English}}.\");\n *\n * prompt.resolve({ language: \"Arabic\" });\n *\n * @example\n * // One-shot seed\n * const prompt = systemPrompt(\"Answer only with JSON matching the schema.\");\n *\n * @example\n * // From a file, read once at construction\n * const prompt = systemPrompt.fromFile(\"./prompts/support-agent.md\");\n *\n * @example\n * // Array form — fully declarative\n * const prompt = systemPrompt([\n *   ai.persona(\"You are Alex.\"),\n *   ai.instruction(\"Always cite sources.\"),\n *   ai.instruction(\"Respond in {{language|English}}.\"),\n * ]);\n */\nexport const systemPrompt: SystemPromptFactory = Object.assign(\n  systemPromptFactory,\n  { fromFile: SystemPrompt.fromFile },\n);\n"],"mappings":";;;;;;;;;;;;;;AAyBA,IAAI,mBAAmB;;;;;;;AAQvB,SAAS,uBACP,OAC+B;CAC/B,OACE,OAAO,UAAU,YACjB,UAAU,QACV,MAAM,QAAS,MAA+B,MAAM,KACpD,OAAQ,MAAgC,YAAY;AAExD;;;;;;AAOA,SAAS,gBAAgB,QAAsC;CAC7D,MAAM,OAAO,OAAO,KAAK;CAEzB,IAAI,MAAM,MACR,OAAO,UAAU,KAAK,MAAM,KAAK,WAAW,GAAG;CAGjD,OAAO,kBAAkB,eAAe,OAAO,KAAK;AACtD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAmDA,IAAa,eAAb,MAAa,aAA6C;CAQxD,AAAO,YACL,AAAgB,SAA+C,CAAC,GAChE,AAAiB,UACjB;EAFgB;EACC;EAEjB,KAAK,KAAK,UAAU;EAOpB,IAAI,UAAU,MACZ,sBAAsB,CAAC,CAAC,SAAS,IAAI;CAEzC;CAUA,AAAO,KACL,MACqD;EACrD,IAAI,SAAS,QACX,OAAO,KAAK;EAGd,OAAO,IAAI,aAAa,KAAK,QAAQ;GAAE,GAAG,KAAK;GAAU,GAAG;EAAK,CAAC;CACpE;;;;;;;;;;;;;;;;;;;;;;;;;;CA2BA,OAAc,SAAS,MAA4B;EACjD,IAAI;EAEJ,IAAI;GACF,WAAW,aAAa,MAAM,MAAM;EACtC,SAAS,OAAO;GACd,MAAM,IAAI,oBACR,sCAAsC,KAAK,MACzC,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK,KAEvD;IAAE,SAAS,EAAE,KAAK;IAAG,OAAO;GAAM,CACpC;EACF;EAEA,OAAO,IAAI,aAAa,CAAC,IAAI,YAAY,QAAQ,CAAC,CAAC;CACrD;;;;;;;;;CAUA,AAAO,QAAQ,OAAuD;EACpE,MAAM,QAAQ,OAAO,UAAU,WAAW,IAAI,QAAQ,KAAK,IAAI;EAC/D,MAAM,gBAAgB,KAAK,OAAO,WAChC,cAAa,UAAU,SAAS,SAClC;EAEA,IAAI,iBAAiB,GAAG;GACtB,MAAM,OAAO,CAAC,GAAG,KAAK,MAAM;GAC5B,KAAK,iBAAiB;GAEtB,OAAO,IAAI,aAAa,IAAI;EAC9B;EAEA,OAAO,IAAI,aAAa,CAAC,OAAO,GAAG,KAAK,MAAM,CAAC;CACjD;;;;;;;CAQA,AAAO,YACL,OACsB;EACtB,MAAM,QAAQ,OAAO,UAAU,WAAW,IAAI,YAAY,KAAK,IAAI;EAEnE,OAAO,IAAI,aAAa,CAAC,GAAG,KAAK,QAAQ,KAAK,CAAC;CACjD;CA6BA,AAAO,MACL,OAOA,GAAG,MAKmB;EAEtB,IAAI,OAAO,UAAU,UAAU;GAC7B,MAAM,UAAU,KAAK;GACrB,MAAM,WAAW,sBAAsB,CAAC,CAAC,IAAI,OAAO,SAAS,WAAW;GAExE,OAAO,KAAK,cAAc,QAAQ;EACpC;EAGA,IAAI,uBAAuB,KAAK,GAC9B,OAAO,KAAK,cAAc,KAAK;EAIjC,MAAM,MAAM,CACV,GAAI,QAAQ,CAAC,KAAK,IAAI,CAAC,GACvB,GAAG,IACL;EAEA,OAAO,KAAK,WAAW,MAAM,GAAG;CAClC;;;;;;CAOA,AAAQ,WACN,OACA,QACsB;EACtB,OAAO,OAAO,QAA8B,QAAQ,UAAU;GAC5D,IAAI,MAAM,SAAS,WACjB,OAAO,OAAO,QAAQ,KAAwB;GAGhD,OAAO,IAAI,aAAa,CAAC,GAAG,OAAO,QAAQ,KAAK,CAAC;EACnD,GAAG,KAAK;CACV;;;;;;;CAQA,AAAQ,cACN,QACsB;EACtB,MAAM,SAAS,KAAK,WAAW,MAAM,OAAO,MAAM;EAMlD,MAAM,eAAe,CAAC,GAHpB,KAAK,UAAU,iBACd,KAAK,UAAU,OAAO,CAAC,gBAAgB,IAAI,CAAC,IAAI,CAAC,IAEX,gBAAgB,MAAM,CAAC;EAIhE,OAAO,IAAI,aAAa,OAAO,QAAQ,EAAE,aAAa,CAAC;CACzD;;;;;;;CAQA,AAAO,QAAQ,cAAqC;EAClD,OAAO,KAAK,OACT,KAAI,UAAS,MAAM,QAAQ,YAAY,CAAC,CAAC,CACzC,KAAK,MAAM,CAAC,CACZ,KAAK;CACV;;;;;;;;CASA,AAAO,SACL,SACiC;EACjC,OAAO,sBAAsB,CAAC,CAAC,SAAS,MAAM,OAAO;CACvD;AACF;AAwBA,SAAS,oBACP,OACA,MACc;CACd,IAAI,UAAU,QACZ,OAAO,IAAI,aAAa,CAAC,GAAG,IAAI;CAGlC,IAAI,OAAO,UAAU,UACnB,OAAO,IAAI,aAAa,CAAC,IAAI,YAAY,KAAK,CAAC,GAAG,IAAI;CAGxD,OAAO,IAAI,aAAa,CAAC,GAAG,KAAK,GAAG,IAAI;AAC1C;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA6CA,MAAa,eAAoC,OAAO,OACtD,qBACA,EAAE,UAAU,aAAa,SAAS,CACpC"}
+{"version":3,"file":"system-prompt.mjs","names":[],"sources":["../../../../../../../@warlock.js/ai/src/system-prompt/system-prompt.ts"],"sourcesContent":["import { readFileSync } from \"node:fs\";\nimport type { Placeholders } from \"../contracts/placeholders.type\";\nimport type {\n  InstructionContract,\n  PersonaContract,\n  RefinedSystemPromptContract,\n  RefinedSystemPromptOptions,\n  SystemPromptBlockContract,\n  SystemPromptContract,\n  SystemPromptMergeOptions,\n  SystemPromptMeta,\n} from \"../contracts/system-prompt.contract\";\nimport { InvalidRequestError } from \"../errors\";\nimport { defaultPromptsManager, promptKey } from \"../prompts/prompts-manager\";\nimport type {\n  PromptValidationResult,\n  PromptsValidateOptions,\n} from \"../prompts/prompts-manager.type\";\nimport { Instruction } from \"./instruction\";\nimport { Persona } from \"./persona\";\nimport { RefinedSystemPrompt } from \"./refined-system-prompt\";\n\n/**\n * Monotonic source of the internal, non-registry display id every\n * `SystemPrompt` carries. Anonymous (unnamed) prompts have nothing else to\n * identify them by; this id never feeds the registry and is never derived from\n * the wall clock, so it stays stable and order-deterministic across a run.\n */\nlet displayIdCounter = 0;\n\n/**\n * Narrow an arbitrary value to a `SystemPromptContract` — true when it exposes\n * the builder surface (`blocks` array + a callable `resolve`). Used by the\n * registry-aware `merge` overload to tell a folded contract from a raw block\n * or a registry name string, robustly across duplicate package copies.\n */\nfunction isSystemPromptContract(\n  value: unknown,\n): value is SystemPromptContract {\n  return (\n    typeof value === \"object\" &&\n    value !== null &&\n    Array.isArray((value as { blocks?: unknown }).blocks) &&\n    typeof (value as { resolve?: unknown }).resolve === \"function\"\n  );\n}\n\n/**\n * Build the deterministic provenance label for a prompt — `name@version` when\n * it is registered, otherwise its internal display id. No random suffixes, so\n * the same source always yields the same `composedFrom` entry.\n */\nfunction provenanceLabel(prompt: SystemPromptContract): string {\n  const meta = prompt.meta();\n\n  if (meta?.name) {\n    return promptKey(meta.name, meta.version ?? \"1\");\n  }\n\n  return prompt instanceof SystemPrompt ? prompt.id : \"anonymous\";\n}\n\n/**\n * Concrete `SystemPromptContract` — an immutable layered prompt builder.\n *\n * **Role.** The top-level composer for a system prompt: it holds an ordered\n * list of typed blocks (persona + instructions) and resolves the whole\n * stack into one final string when the agent is about to call the model.\n *\n * **Responsibility.**\n * - Owns: the ordered `blocks` list and the block-join rules (insertion\n *   order, blank-line separator, trim).\n * - Does NOT own: how any individual block is rendered (delegated to each\n *   block's `resolve()`), the placeholder syntax (delegated to\n *   `renderPlaceholders`), or any knowledge of the agent, model, or\n *   session consuming the resolved text.\n *\n * Blocks are discriminated by a string `type` tag (`\"persona\"` /\n * `\"instruction\"`) rather than `instanceof`, so user-supplied blocks that\n * implement `SystemPromptBlockContract` interoperate seamlessly with blocks\n * built via `ai.persona()` / `ai.instruction()` — even across duplicate\n * package copies or bundler scope boundaries.\n *\n * The builder is **immutable** — every `.persona()` / `.instruction()`\n * call returns a fresh `SystemPrompt` instance sharing nothing mutable\n * with its parent. This makes forking a base prompt into specialized\n * variants a safe, side-effect-free operation.\n *\n * Users construct via the `ai.systemPrompt()` factory — `new SystemPrompt()`\n * is not the public API (see §4.2 of code-style.md). Modeled as a class so\n * that methods live on the prototype (one copy shared across every forked\n * instance) and downstream code can branch via `instanceof SystemPrompt`.\n *\n * @example\n * // Chainable form\n * const alex = ai.persona(\"You are Alex, a TypeScript expert.\");\n * const replyIn = ai.instruction(\"Respond in {{language|English}}.\");\n *\n * const base = ai.systemPrompt().persona(alex).instruction(replyIn);\n * const arabicVariant = base.instruction(\"Prefer Arabic comments.\");\n *\n * base.resolve({ language: \"English\" });\n * arabicVariant.resolve({ language: \"Arabic\" });\n *\n * @example\n * // Array form — insertion order is preserved exactly\n * const prompt = ai.systemPrompt([\n *   ai.persona(\"You are Alex, a TypeScript expert.\"),\n *   ai.instruction(\"Respond in {{language|English}}.\"),\n * ]);\n */\nexport class SystemPrompt implements SystemPromptContract {\n  /**\n   * Internal, non-registry id for display / provenance. Stable for the life of\n   * the instance; sourced from a monotonic counter, never the wall clock.\n   * Anonymous prompts are identified solely by this id.\n   */\n  public readonly id: string;\n\n  public constructor(\n    public readonly blocks: readonly SystemPromptBlockContract[] = [],\n    private readonly metaData?: SystemPromptMeta,\n  ) {\n    this.id = `prompt#${displayIdCounter++}`;\n\n    // Auto-register the moment a builder acquires a name — whether through the\n    // `systemPrompt(input, { name })` factory or a `.meta({ name })` rename.\n    // Forks built by `persona()` / `instruction()` / `merge()` deliberately\n    // drop the name (they pass no meta), so they stay anonymous and never land\n    // in the registry unless explicitly re-named.\n    if (metaData?.name) {\n      defaultPromptsManager().register(this);\n    }\n  }\n\n  /**\n   * Read the current metadata snapshot (no argument) or derive a renamed\n   * builder (with `meta`). The accessor returns `undefined` for an anonymous\n   * prompt; the updater shallow-merges `meta` onto the current metadata and\n   * returns a fresh builder. Naming the result registers it in `ai.prompts`.\n   */\n  public meta(): SystemPromptMeta | undefined;\n  public meta(meta: SystemPromptMeta): SystemPromptContract;\n  public meta(\n    meta?: SystemPromptMeta,\n  ): SystemPromptMeta | undefined | SystemPromptContract {\n    if (meta === undefined) {\n      return this.metaData;\n    }\n\n    return new SystemPrompt(this.blocks, { ...this.metaData, ...meta });\n  }\n\n  /**\n   * Build a system prompt by reading the file at `path` once, synchronously,\n   * at construction time. The file's UTF-8 contents seed a single instruction\n   * block — the same semantics as the string-seed form of `systemPrompt()` —\n   * so placeholders inside the file (`{{language|English}}`) resolve at\n   * `resolve()` time and the result can be forked with further\n   * `.persona()` / `.instruction()` calls.\n   *\n   * One-shot by design: the file is read exactly once here, never re-read on\n   * `resolve()`. Reads are synchronous so the call stays a drop-in for the\n   * synchronous `systemPrompt()` factory and the synchronous `resolve()` API.\n   *\n   * Throws `InvalidRequestError` when the file cannot be read (missing path,\n   * permission denied) — surfacing the underlying cause so a typo in the\n   * prompt path fails loudly at construction instead of silently producing an\n   * empty prompt.\n   *\n   * @param path - Filesystem path to the prompt template file.\n   *\n   * @example\n   * const prompt = SystemPrompt.fromFile(\"./prompts/support-agent.md\");\n   *\n   * const localized = prompt.instruction(\"Respond in {{language|English}}.\");\n   * localized.resolve({ language: \"Arabic\" });\n   */\n  public static fromFile(path: string): SystemPrompt {\n    let contents: string;\n\n    try {\n      contents = readFileSync(path, \"utf8\");\n    } catch (error) {\n      throw new InvalidRequestError(\n        `Failed to read system prompt file \"${path}\" — ${\n          error instanceof Error ? error.message : String(error)\n        }`,\n        { context: { path }, cause: error },\n      );\n    }\n\n    return new SystemPrompt([new Instruction(contents)]);\n  }\n\n  /**\n   * Return a new builder with the persona block set. If a persona already\n   * exists it's replaced in place (preserving its position in `blocks`);\n   * otherwise the new persona is prepended so persona-first remains the\n   * default for chain-built prompts. Accepts either raw text (auto-wrapped\n   * via `new Persona`) or an existing `PersonaContract` instance for reuse\n   * across prompts.\n   */\n  public persona(value: PersonaContract | string): SystemPromptContract {\n    const block = typeof value === \"string\" ? new Persona(value) : value;\n    const existingIndex = this.blocks.findIndex(\n      candidate => candidate.type === \"persona\",\n    );\n\n    if (existingIndex >= 0) {\n      const next = [...this.blocks];\n      next[existingIndex] = block;\n\n      return new SystemPrompt(next) as this;\n    }\n\n    return new SystemPrompt([block, ...this.blocks]);\n  }\n\n  /**\n   * Return a new builder with the given instruction appended. Instructions\n   * render in insertion order. Accepts either raw text (auto-wrapped via\n   * `new Instruction`) or an existing `InstructionContract` instance for\n   * cross-prompt reuse.\n   */\n  public instruction(\n    value: InstructionContract | string,\n  ): SystemPromptContract {\n    const block = typeof value === \"string\" ? new Instruction(value) : value;\n\n    return new SystemPrompt([...this.blocks, block]);\n  }\n\n  /**\n   * Fold predefined blocks, another prompt contract, or a registered prompt\n   * name into this builder. Three forms share one method:\n   *\n   * - `merge(...blocks)` — N pre-built `ai.persona()` / `ai.instruction()`\n   *   blocks. A `persona` block sets/replaces the single, leading persona;\n   *   every other block appends in order. `base.merge(reviewer, style, lang)`\n   *   equals `base.persona(reviewer).instruction(style).instruction(lang)`.\n   * - `merge(contract)` — another prompt; its blocks fold in (persona\n   *   replaces, instructions append) and `meta.composedFrom` records the\n   *   provenance of both sides.\n   * - `merge(name, { fromVersion })` — a prompt resolved from `ai.prompts`\n   *   (latest version unless `fromVersion` selects another); throws\n   *   `InvalidRequestError` when the name / version is unregistered.\n   *\n   * Immutable — the original builder is untouched; passing zero blocks returns\n   * an equivalent builder. The folded result is anonymous (no `name`), so it\n   * is never auto-registered even though it carries `composedFrom` provenance.\n   */\n  public merge(\n    ...blocks: readonly SystemPromptBlockContract[]\n  ): SystemPromptContract;\n  public merge(source: SystemPromptContract): SystemPromptContract;\n  public merge(\n    name: string,\n    options?: SystemPromptMergeOptions,\n  ): SystemPromptContract;\n  public merge(\n    first?:\n      | SystemPromptBlockContract\n      | SystemPromptContract\n      | string,\n    // `undefined` is part of the element union so the `merge(name, options?)`\n    // overload's optional trailing `options?` (i.e. `… | undefined`) stays\n    // assignable to this implementation signature.\n    ...rest: readonly (\n      | SystemPromptBlockContract\n      | SystemPromptMergeOptions\n      | undefined\n    )[]\n  ): SystemPromptContract {\n    // Registry-name form: resolve from ai.prompts at the chosen version.\n    if (typeof first === \"string\") {\n      const options = rest[0] as SystemPromptMergeOptions | undefined;\n      const resolved = defaultPromptsManager().get(first, options?.fromVersion);\n\n      return this.mergeContract(resolved);\n    }\n\n    // Contract form: fold another prompt's blocks + record provenance.\n    if (isSystemPromptContract(first)) {\n      return this.mergeContract(first);\n    }\n\n    // Variadic block form (the original behavior).\n    const all = [\n      ...(first ? [first] : []),\n      ...rest,\n    ] as readonly SystemPromptBlockContract[];\n\n    return this.foldBlocks(this, all);\n  }\n\n  /**\n   * Fold an ordered list of blocks onto a starting prompt: persona blocks\n   * set/replace the single leading persona; every other block appends in\n   * order. The shared core of the variadic-block `merge` and the contract fold.\n   */\n  private foldBlocks(\n    start: SystemPromptContract,\n    blocks: readonly SystemPromptBlockContract[],\n  ): SystemPromptContract {\n    return blocks.reduce<SystemPromptContract>((prompt, block) => {\n      if (block.type === \"persona\") {\n        return prompt.persona(block as PersonaContract);\n      }\n\n      return new SystemPrompt([...prompt.blocks, block]);\n    }, start);\n  }\n\n  /**\n   * Fold another prompt contract into this one (persona replaces, instructions\n   * append) and stamp the deterministic `composedFrom` provenance — this\n   * prompt's existing provenance (or its own label) followed by the folded\n   * source's label. The result is anonymous so it never auto-registers.\n   */\n  private mergeContract(\n    source: SystemPromptContract,\n  ): SystemPromptContract {\n    const folded = this.foldBlocks(this, source.blocks);\n\n    const baseProvenance =\n      this.metaData?.composedFrom ??\n      (this.metaData?.name ? [provenanceLabel(this)] : []);\n\n    const composedFrom = [...baseProvenance, provenanceLabel(source)];\n\n    // Carry forward only provenance — never the name — so the merged result is\n    // a fresh anonymous prompt (immutable rename = new key; original stays).\n    return new SystemPrompt(folded.blocks, { composedFrom });\n  }\n\n  /**\n   * Resolve every block against the placeholder map, join the results with\n   * blank-line separators (in insertion order), and trim. Returns an empty\n   * string when no blocks are present — callers treat that as \"no system\n   * message\".\n   */\n  public resolve(placeholders?: Placeholders): string {\n    return this.blocks\n      .map(block => block.resolve(placeholders))\n      .join(\"\\n\\n\")\n      .trim();\n  }\n\n  /**\n   * Validate this prompt via the process-wide `ai.prompts` manager — sugar for\n   * `ai.prompts.validate(this, options)`. Runs the deterministic placeholder\n   * check and, when `options.judge` is supplied, the Nova-safe LLM-as-judge\n   * pass. Never throws on a judge failure; `ok` tracks the deterministic\n   * verdict alone.\n   */\n  public validate(\n    options?: PromptsValidateOptions,\n  ): Promise<PromptValidationResult> {\n    return defaultPromptsManager().validate(this, options);\n  }\n\n  /**\n   * Derive the compiled form of this prompt — a lazy wrapper that rewrites\n   * the human-authored text into a model-optimized version on first use,\n   * pins the result, and serves the pin thereafter. See\n   * {@link RefinedSystemPromptContract} for the full semantics (lockfile\n   * pinning, placeholder parity, advisory fallback, `refine()` /\n   * `refinePrompt()`).\n   *\n   * The wrapper's collaborators are injected here rather than imported by\n   * `refined-system-prompt.ts` — importing this module (or the prompts\n   * manager) back from there would close an import cycle.\n   */\n  public refined(\n    options: RefinedSystemPromptOptions,\n  ): RefinedSystemPromptContract {\n    return new RefinedSystemPrompt(this, options, {\n      buildPrompt: (blocks, meta) => new SystemPrompt([...blocks], meta),\n      validatePrompt: (target, validateOptions) =>\n        defaultPromptsManager().validate(target, validateOptions),\n    });\n  }\n}\n\n/**\n * Public factory for `SystemPrompt`, callable directly or via its\n * `fromFile` static. Exists as a named interface so the callable signature\n * and the `fromFile` attachment travel together as one public type.\n */\nexport interface SystemPromptFactory {\n  (\n    input?: string | ReadonlyArray<SystemPromptBlockContract>,\n    meta?: SystemPromptMeta,\n  ): SystemPrompt;\n\n  /**\n   * Build a system prompt from a file read once at construction. Delegates\n   * to {@link SystemPrompt.fromFile}, so `ai.systemPrompt.fromFile(path)` and\n   * `SystemPrompt.fromFile(path)` behave identically.\n   *\n   * @example\n   * const prompt = ai.systemPrompt.fromFile(\"./prompts/support-agent.md\");\n   */\n  fromFile(path: string): SystemPrompt;\n}\n\nfunction systemPromptFactory(\n  input?: string | ReadonlyArray<SystemPromptBlockContract>,\n  meta?: SystemPromptMeta,\n): SystemPrompt {\n  if (input === undefined) {\n    return new SystemPrompt([], meta);\n  }\n\n  if (typeof input === \"string\") {\n    return new SystemPrompt([new Instruction(input)], meta);\n  }\n\n  return new SystemPrompt([...input], meta);\n}\n\n/**\n * Create a new immutable system-prompt builder.\n *\n * **Role.** Public factory for `SystemPrompt` — keeps user-facing code\n * free of `new` and consistent with `ai.tool()`, `ai.agent()`,\n * `ai.persona()`, `ai.instruction()`.\n *\n * Input forms:\n * - No argument → empty builder, chain `.persona()` / `.instruction()`\n * - Single string → seeded with one instruction for quick one-shot prompts\n * - Array of blocks → used verbatim, preserving insertion order\n * - `.fromFile(path)` → seeded from a file read once at construction\n *\n * Pass a second `meta` argument to name the prompt — a named prompt\n * auto-registers in `ai.prompts` under `name@version` (version defaults to the\n * next integer). Forks (`.persona()`, `.instruction()`, `.merge()`) are\n * anonymous unless re-named via `.meta({ name })`.\n *\n * @example\n * // Composed builder\n * const prompt = systemPrompt()\n *   .persona(\"You are Alex, a senior TypeScript engineer.\")\n *   .instruction(\"Always include working code examples.\")\n *   .instruction(\"Respond in {{language|English}}.\");\n *\n * prompt.resolve({ language: \"Arabic\" });\n *\n * @example\n * // One-shot seed\n * const prompt = systemPrompt(\"Answer only with JSON matching the schema.\");\n *\n * @example\n * // From a file, read once at construction\n * const prompt = systemPrompt.fromFile(\"./prompts/support-agent.md\");\n *\n * @example\n * // Array form — fully declarative\n * const prompt = systemPrompt([\n *   ai.persona(\"You are Alex.\"),\n *   ai.instruction(\"Always cite sources.\"),\n *   ai.instruction(\"Respond in {{language|English}}.\"),\n * ]);\n */\nexport const systemPrompt: SystemPromptFactory = Object.assign(\n  systemPromptFactory,\n  { fromFile: SystemPrompt.fromFile },\n);\n"],"mappings":";;;;;;;;;;;;;;;AA4BA,IAAI,mBAAmB;;;;;;;AAQvB,SAAS,uBACP,OAC+B;CAC/B,OACE,OAAO,UAAU,YACjB,UAAU,QACV,MAAM,QAAS,MAA+B,MAAM,KACpD,OAAQ,MAAgC,YAAY;AAExD;;;;;;AAOA,SAAS,gBAAgB,QAAsC;CAC7D,MAAM,OAAO,OAAO,KAAK;CAEzB,IAAI,MAAM,MACR,OAAO,UAAU,KAAK,MAAM,KAAK,WAAW,GAAG;CAGjD,OAAO,kBAAkB,eAAe,OAAO,KAAK;AACtD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAmDA,IAAa,eAAb,MAAa,aAA6C;CAQxD,AAAO,YACL,AAAgB,SAA+C,CAAC,GAChE,AAAiB,UACjB;EAFgB;EACC;EAEjB,KAAK,KAAK,UAAU;EAOpB,IAAI,UAAU,MACZ,sBAAsB,CAAC,CAAC,SAAS,IAAI;CAEzC;CAUA,AAAO,KACL,MACqD;EACrD,IAAI,SAAS,QACX,OAAO,KAAK;EAGd,OAAO,IAAI,aAAa,KAAK,QAAQ;GAAE,GAAG,KAAK;GAAU,GAAG;EAAK,CAAC;CACpE;;;;;;;;;;;;;;;;;;;;;;;;;;CA2BA,OAAc,SAAS,MAA4B;EACjD,IAAI;EAEJ,IAAI;GACF,WAAW,aAAa,MAAM,MAAM;EACtC,SAAS,OAAO;GACd,MAAM,IAAI,oBACR,sCAAsC,KAAK,MACzC,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK,KAEvD;IAAE,SAAS,EAAE,KAAK;IAAG,OAAO;GAAM,CACpC;EACF;EAEA,OAAO,IAAI,aAAa,CAAC,IAAI,YAAY,QAAQ,CAAC,CAAC;CACrD;;;;;;;;;CAUA,AAAO,QAAQ,OAAuD;EACpE,MAAM,QAAQ,OAAO,UAAU,WAAW,IAAI,QAAQ,KAAK,IAAI;EAC/D,MAAM,gBAAgB,KAAK,OAAO,WAChC,cAAa,UAAU,SAAS,SAClC;EAEA,IAAI,iBAAiB,GAAG;GACtB,MAAM,OAAO,CAAC,GAAG,KAAK,MAAM;GAC5B,KAAK,iBAAiB;GAEtB,OAAO,IAAI,aAAa,IAAI;EAC9B;EAEA,OAAO,IAAI,aAAa,CAAC,OAAO,GAAG,KAAK,MAAM,CAAC;CACjD;;;;;;;CAQA,AAAO,YACL,OACsB;EACtB,MAAM,QAAQ,OAAO,UAAU,WAAW,IAAI,YAAY,KAAK,IAAI;EAEnE,OAAO,IAAI,aAAa,CAAC,GAAG,KAAK,QAAQ,KAAK,CAAC;CACjD;CA6BA,AAAO,MACL,OAOA,GAAG,MAKmB;EAEtB,IAAI,OAAO,UAAU,UAAU;GAC7B,MAAM,UAAU,KAAK;GACrB,MAAM,WAAW,sBAAsB,CAAC,CAAC,IAAI,OAAO,SAAS,WAAW;GAExE,OAAO,KAAK,cAAc,QAAQ;EACpC;EAGA,IAAI,uBAAuB,KAAK,GAC9B,OAAO,KAAK,cAAc,KAAK;EAIjC,MAAM,MAAM,CACV,GAAI,QAAQ,CAAC,KAAK,IAAI,CAAC,GACvB,GAAG,IACL;EAEA,OAAO,KAAK,WAAW,MAAM,GAAG;CAClC;;;;;;CAOA,AAAQ,WACN,OACA,QACsB;EACtB,OAAO,OAAO,QAA8B,QAAQ,UAAU;GAC5D,IAAI,MAAM,SAAS,WACjB,OAAO,OAAO,QAAQ,KAAwB;GAGhD,OAAO,IAAI,aAAa,CAAC,GAAG,OAAO,QAAQ,KAAK,CAAC;EACnD,GAAG,KAAK;CACV;;;;;;;CAQA,AAAQ,cACN,QACsB;EACtB,MAAM,SAAS,KAAK,WAAW,MAAM,OAAO,MAAM;EAMlD,MAAM,eAAe,CAAC,GAHpB,KAAK,UAAU,iBACd,KAAK,UAAU,OAAO,CAAC,gBAAgB,IAAI,CAAC,IAAI,CAAC,IAEX,gBAAgB,MAAM,CAAC;EAIhE,OAAO,IAAI,aAAa,OAAO,QAAQ,EAAE,aAAa,CAAC;CACzD;;;;;;;CAQA,AAAO,QAAQ,cAAqC;EAClD,OAAO,KAAK,OACT,KAAI,UAAS,MAAM,QAAQ,YAAY,CAAC,CAAC,CACzC,KAAK,MAAM,CAAC,CACZ,KAAK;CACV;;;;;;;;CASA,AAAO,SACL,SACiC;EACjC,OAAO,sBAAsB,CAAC,CAAC,SAAS,MAAM,OAAO;CACvD;;;;;;;;;;;;;CAcA,AAAO,QACL,SAC6B;EAC7B,OAAO,IAAI,oBAAoB,MAAM,SAAS;GAC5C,cAAc,QAAQ,SAAS,IAAI,aAAa,CAAC,GAAG,MAAM,GAAG,IAAI;GACjE,iBAAiB,QAAQ,oBACvB,sBAAsB,CAAC,CAAC,SAAS,QAAQ,eAAe;EAC5D,CAAC;CACH;AACF;AAwBA,SAAS,oBACP,OACA,MACc;CACd,IAAI,UAAU,QACZ,OAAO,IAAI,aAAa,CAAC,GAAG,IAAI;CAGlC,IAAI,OAAO,UAAU,UACnB,OAAO,IAAI,aAAa,CAAC,IAAI,YAAY,KAAK,CAAC,GAAG,IAAI;CAGxD,OAAO,IAAI,aAAa,CAAC,GAAG,KAAK,GAAG,IAAI;AAC1C;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA6CA,MAAa,eAAoC,OAAO,OACtD,qBACA,EAAE,UAAU,aAAa,SAAS,CACpC"}

package/llms-full.txt

@@ -2439,7 +2439,7 @@ A bare `ai.supervisor()` / `ai.workflow()` uses a `snapshotStore` for `resume(ru
 
 ---
 name: manage-prompts
-description: 'Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`; ''register a prompt by name'', ''resolve a prompt by name@version or tag'', ''pin a production tag to a prompt version'', ''diff two prompt versions'', ''export / import the prompt registry'', ''validate a prompt for missing placeholders'', ''merge a registered prompt into another''; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.'
+description: 'Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`, `criteria`; ''register a prompt by name'', ''resolve a prompt by name@version or tag'', ''pin a production tag to a prompt version'', ''diff two prompt versions'', ''export / import the prompt registry'', ''validate a prompt for missing placeholders'', ''validate a prompt against my own criteria / rules'', ''merge a registered prompt into another''; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.'
 ---
 
 # `ai.prompts` — the unified prompt registry
@@ -2541,6 +2541,11 @@ const report = await ai.prompts.validate("support", {
   placeholders: { product: "Warlock" }, // values you intend to supply
   declare: ["language"],                  // extra keys to treat as known
   judge: judgeModel,                      // optional — turns on the LLM-as-judge pass
+  criteria: [                             // optional — YOUR rules, replaces the built-in rubric
+    "Addresses the user by {{name}}",
+    "Never gives medical advice",
+    "Stays under 200 words",
+  ],
 });
 
 report.ok;       // true iff no required placeholder is missing (DETERMINISTIC verdict alone)
@@ -2551,6 +2556,7 @@ report.issues;   // advisory judge reasons / a degrade note — present only whe
 
 - **Always** runs the deterministic check: every `{{key}}` with no inline default that is neither supplied (`placeholders`), declared (`declare`), nor in the prompt's `meta.required` lands in `missing`; `ok` is `true` iff `missing` is empty.
 - **`judge`** adds a **Nova-safe** LLM-as-judge quality pass — it **never throws** and degrades to an `issues` note (leaving `score` undefined) on failure, so a flaky judge can **never flip `ok`**.
+- **`criteria`** (a string or a list of short rules) grades the prompt against **your own rules** instead of the built-in quality rubric — `score` / `issues` then reflect your criteria (a failed rule is named in `issues`). Only used when `judge` is also set; folded into the `judgeCache` key so different rules re-run. Still advisory — never flips `ok`.
 - **`target`** is a registered name (or `name@selector`), a `SystemPromptContract` instance, or a raw prompt string.
 - **`judgeCache`** (per-call or via the `prompts({ judgeCache })` factory option) memoizes judge verdicts by a content hash of the resolved body + the judge model id — a structural `{ get, set }` subset of `@warlock.js/cache`'s `CacheDriver`, so the cache package stays a strictly **optional** peer.
 
@@ -2621,6 +2627,7 @@ If you only ever called `ai.prompt({ ... })` and used the returned registry, **n
 ## See also
 
 - [`@warlock.js/ai/write-system-prompt/SKILL.md`](@warlock.js/ai/write-system-prompt/SKILL.md) — the `systemPrompt()` / `persona()` / `instruction()` builder, `.meta()`, and `merge()` this registry stores and composes
+- [`@warlock.js/ai/refine-prompts/SKILL.md`](@warlock.js/ai/refine-prompts/SKILL.md) — `systemPrompt().refined({ model, criteria, store })`, the prompt compiler; register its `refinePrompt()` output as a next version to `diff` original vs refined
 - [`@warlock.js/ai/eval-datasets-and-ci/SKILL.md`](@warlock.js/ai/eval-datasets-and-ci/SKILL.md) — the eval `judge` scorer `validate()`'s LLM pass reuses
 - [`@warlock.js/ai/run-ai-agent/SKILL.md`](@warlock.js/ai/run-ai-agent/SKILL.md) — wiring a resolved prompt into an agent, plus the judge-safe agent preset (`ai.agent.judge`)
 
@@ -3411,6 +3418,101 @@ VCR composes **below** `ai.fallbackModel` and works with any adapter. Wrap the l
 - [`@warlock.js/ai/pick-ai-provider/SKILL.md`](@warlock.js/ai/pick-ai-provider/SKILL.md) — the adapters whose models VCR wraps
 
 
+## refine-prompts  `@warlock.js/ai/refine-prompts/SKILL.md`
+
+---
+name: refine-prompts
+description: 'Prompt compiler — systemPrompt(...).refined({ model, criteria, store }): humans keep writing human prompt text (dev code or admin-panel textareas); the refined wrapper lazily rewrites it into a model-optimized version via a refiner model on first agent use, pins the result (lockfile posture — recompiled ONLY when the source text, refiner model, criteria, or built-in recipe version change, never silently over time), and serves the pin thereafter. Explicit surfaces: await refined.refine() → the compiled template STRING (placeholders intact — for admin routes, previews, boot warmup, CI; throws PromptRefinementError on failure) and await refined.refinePrompt() → a composable SystemPromptContract with meta.refinedFrom / meta.refinerModel provenance (register it as a next version to unlock ai.prompts.diff review). Placeholder parity is machine-enforced (the exact {{placeholder}} set must survive or the rewrite is rejected after one repair re-ask); the lazy agent path NEVER throws — on refiner failure it warns once and serves the original. store is a structural { get, set } (any @warlock.js/cache CacheDriver); omitted ⇒ the pin lives on the instance for the process lifetime. Triggers: `refined`, `refine`, `refinePrompt`, `materialize`, `RefinedSystemPromptContract`, `RefinedSystemPromptOptions`, `RefinedPromptStoreLike`, `PromptRefineOptions`, `PromptRefinementError`, `refinedFrom`, `refinerModel`, `fresh`, `prompt-refiner`, ''refine a prompt'', ''compile a prompt'', ''optimize a system prompt'', ''rewrite my prompt to be AI-friendly'', ''admin-written prompts'', ''prompt refinement store''; typical import `import { ai } from "@warlock.js/ai"`. Skip: registry operations (register / resolve / tag / diff / validate) — `@warlock.js/ai/manage-prompts/SKILL.md`; composing prompts from persona + instruction blocks — `@warlock.js/ai/write-system-prompt/SKILL.md`; grading a prompt against rules without rewriting it — validate({ criteria }) in `@warlock.js/ai/manage-prompts/SKILL.md`.'
+---
+
+# `systemPrompt().refined()` — the prompt compiler
+
+Humans write prompts as human text; models perform better on structured, model-tuned phrasing. `.refined({ model, criteria, store })` turns any `SystemPrompt` into a **lazily-compiled artifact**: the first agent use rewrites the raw source template through the refiner `model`, pins the result, and every later use serves the pin. The human text stays the editing surface forever — the refined text is a derived artifact, like a lockfile.
+
+```ts
+import { ai } from "@warlock.js/ai";
+
+const support = ai
+  .systemPrompt(
+    [ai.persona("You are a friendly assistant."), ai.instruction("Help {{name}} with orders.")],
+    { name: "support" },
+  )
+  .refined({ model: refinerModel, store: myCacheDriver });
+
+// Lazy: compiles on the first run, serves the pin afterwards.
+const agent = ai.agent({ model, systemPrompt: support });
+
+// Explicit: compile now — admin routes, previews, boot warmup, CI.
+const text = await support.refine();          // the compiled template STRING
+const prompt = await support.refinePrompt();  // a composable SystemPromptContract
+```
+
+## The four trust rules
+
+1. **Lockfile posture.** The pin key hashes the recipe version + refiner model + `criteria` + source template — any input change compiles fresh; an unchanged input NEVER recompiles (no TTL, no silent drift). `store` is a **store, not a cache**.
+2. **Prose, never contract.** The exact `{{placeholder}}` set (name **and** `|default`) must survive the rewrite verbatim — checked mechanically; a parity break gets ONE repair re-ask, then the rewrite is rejected. The compiled text is still a **template**: placeholders resolve per call as usual.
+3. **Advisory with fallback.** The lazy agent path never throws: a refiner failure warns once (`[warlock-ai] …`) and serves the ORIGINAL prompt — the human text is always a valid prompt. After **3** failed attempts the lazy path stops retrying for the instance lifetime (no per-run refiner latency from a broken key/provider); the explicit `refine()` / `refinePrompt()` stay live — they **throw** `PromptRefinementError` (`error.reason`: `"model"` / `"parity"` / `"empty"`) and a later success re-arms the pin for everyone.
+4. **Reviewable.** `refine()` exposes the compiled text; `refinePrompt()` makes it a first-class prompt with provenance.
+
+## `refine(options?)` — the explicit string surface
+
+```ts
+const text = await support.refine();                 // store-first; pins on first compile
+const another = await support.refine({ fresh: true }); // skip the pin, new take, re-pins
+```
+
+Expose it via a route for an admin **preview / approve** flow — the admin sees original vs refined, and the call itself warms the pin so the next agent run pays nothing. Also the boot-warmup / CI-compile surface.
+
+## `refinePrompt(options?)` — the composable surface
+
+```ts
+const compiled = await support.refinePrompt();
+
+compiled.blocks;                 // one instruction block = the refined template
+compiled.meta()?.refinedFrom;    // "support@1" (or "anonymous")
+compiled.meta()?.refinerModel;   // "anthropic:claude-sonnet-4-5"
+compiled.meta()?.required;       // carried from the source — contract preserved
+```
+
+It never auto-registers (no `name` — registry versions stay human-intentional). Register it deliberately to unlock the review flow:
+
+```ts
+compiled.meta({ name: "support" });          // registers as support@<next>
+ai.prompts.diff("support", "1", "2");        // original vs refined, block by block
+```
+
+## Options
+
+- **`model`** (required) — the refiner `ModelContract`. The call runs as a one-shot `"prompt-refiner"` agent, so usage/cost surface through the standard report/observer machinery.
+- **`criteria`** — a string or list of rules the rewrite MUST satisfy, on top of the built-in recipe. Same word and shape as `validate({ criteria })`: *validate grades against criteria; refined rewrites against them*. Folded into the pin key — new rules compile fresh.
+- **`store`** — structural `{ get, set }` (`RefinedPromptStoreLike`; any `@warlock.js/cache` `CacheDriver` satisfies it — the cache package stays an optional peer). Share a redis/pg-backed driver so ONE process pays each compilation and the fleet reads the pin. Omitted ⇒ the pin lives on the wrapper instance for the process lifetime. A pinned value that fails the parity check (corrupt / tampered store) is treated as a miss and recompiled.
+
+## What compiles where — the lazy boundary
+
+The lazy compile hook rides the **agent path** (`ai.agent` execute/stream, and everything built on it — supervisors' member agents, planner steps, eval, `spawnSubAgent`, `serve`). Prompts resolved **synchronously at factory time** — `ai.planner({ systemPrompt })` / `ai.router({ systemPrompt })` prefixes, a supervisor's own `systemPrompt` / `goal`, and `ai.prompts.resolve()` — use the ORIGINAL text unless you pre-warm:
+
+```ts
+await refined.refine();                       // warm the pin at boot…
+const planner = ai.planner({ systemPrompt: refined, ... }); // …then factories see it? NO —
+```
+
+Factory-time resolution reads whatever is pinned **at that moment** — so warm BEFORE constructing the factory, or pass `await refined.refinePrompt()` instead (an already-compiled plain prompt).
+
+## Chaining and identity
+
+- `refined.meta()` reads the SOURCE meta — agent reports stamp the source `name@version`, so observability groups by the prompt you authored.
+- `.persona()` / `.instruction()` / `.merge()` / `.meta({...})` derive a NEW source and re-wrap it with the same refinement options — editing a compiled prompt invalidates its pin naturally (new source ⇒ new key).
+- `refined.source` is always the original builder; `refined.resolve(placeholders)` serves the compiled text once pinned, the original before.
+- **Register the source or the `refinePrompt()` output — not the wrapper itself.** The wrapper's `blocks` flip from source to compiled text on materialization, so `ai.prompts.register(wrapper)` would fingerprint whatever is pinned at call time (and a re-register after the flip throws on the content mismatch).
+- `validate()` on the wrapper validates what it currently serves — pair `refined` with `validate({ criteria, judge })` to lint the compiled text, and with `agent.eval` (original vs refined on a dataset) to PROVE the rewrite helps before trusting it.
+
+## See also
+
+- [`@warlock.js/ai/manage-prompts/SKILL.md`](@warlock.js/ai/manage-prompts/SKILL.md) — the registry (`name@version`, tags, `diff`, `validate({ criteria })`) the review flow rides on
+- [`@warlock.js/ai/write-system-prompt/SKILL.md`](@warlock.js/ai/write-system-prompt/SKILL.md) — the `systemPrompt()` builder `.refined()` extends
+- [`@warlock.js/ai/eval-datasets-and-ci/SKILL.md`](@warlock.js/ai/eval-datasets-and-ci/SKILL.md) — measure original vs refined behaviour on a dataset
+
+
 ## run-ai-agent  `@warlock.js/ai/run-ai-agent/SKILL.md`
 
 ---
@@ -5560,6 +5662,7 @@ Three distinct prompts, one common foundation. Base is immutable — safe to sha
 ## See also
 
 - [`@warlock.js/ai/manage-prompts/SKILL.md`](@warlock.js/ai/manage-prompts/SKILL.md) — the `ai.prompts` registry these named prompts auto-register into (resolve / version / tag / diff / export / validate)
+- [`@warlock.js/ai/refine-prompts/SKILL.md`](@warlock.js/ai/refine-prompts/SKILL.md) — `.refined({ model, criteria, store })`, the prompt compiler: lazily rewrite this builder into a model-optimized version, pinned like a lockfile
 - [`@warlock.js/ai/run-ai-agent/SKILL.md`](@warlock.js/ai/run-ai-agent/SKILL.md) — `systemPrompt` on factory + per-call override
 - [`@warlock.js/ai/run-ai-workflow/SKILL.md`](@warlock.js/ai/run-ai-workflow/SKILL.md) — per-step agent references inherit their own system prompt
 

package/llms.txt

@@ -23,12 +23,13 @@
 - [handle-ai-errors](@warlock.js/ai/handle-ai-errors/SKILL.md): Typed AIError hierarchy with stable code strings + coarse category for retry-policy dispatch. execute() never throws — errors surface via result.error (the sole exception: OrchestratorConfigError throws at construction). Triggers: `AIError`, `ProviderRateLimitError`, `ProviderAuthError`, `ContextLengthExceededError`, `ContentFilterError`, `SchemaValidationError`, `ToolExecutionError`, `WorkflowDriftError`, `SupervisorDriftError`, `SupervisorFailedError`, `SupervisorRoutingError`, `OrchestratorFailedError`, `OrchestratorDriftError`, `OrchestratorConfigError`, `OrchestratorCancelledError`, `PlannerFailedError`, `PlannerPlanInvalidError`, `PlannerCancelledError`, `BudgetExceededError`, `GuardrailViolationError`, `error.code`, `error.category`; 'handle ai error', 'retry on rate limit', 'branch on error code', 'ORCHESTRATOR_DRIFT', 'PLANNER_PLAN_INVALID', 'build fallback ladder'; typical import `import { AIError } from "@warlock.js/ai"`. Skip: log surfacing — `@warlock.js/ai/log-ai-calls/SKILL.md`; native `try / catch` on raw `openai`.
 - [log-ai-calls](@warlock.js/ai/log-ai-calls/SKILL.md): Framework logging delegated to @warlock.js/logger — every primitive emits via the log singleton, configure channels / levels / redaction once at boot. Four-arg call convention (module, action, message, context). Triggers: `log.configure`, `log.setMinLevel`, `log.setChannels`, `ConsoleLog`, `FileLog`, `LogChannel`, `redact.paths`, `ai.agent.<name>` / `ai.workflow.<name>` / `ai.supervisor.<name>` modules; 'configure ai logging', 'mask prompts in logs', 'silence logs in tests', 'capture log entries'; typical import `import { log } from "@warlock.js/logger"`. Skip: error hierarchy — `@warlock.js/ai/handle-ai-errors/SKILL.md`; competing libs `pino`, `winston`, `console.log`.
 - [manage-ai-stores](@warlock.js/ai/manage-ai-stores/SKILL.md): Durable orchestrator stores — ai.checkpoint.{memory,pg,redis}() for cross-turn SESSION STATE and ai.snapshot.{memory,pg,redis}() for in-flight SUPERVISOR/WORKFLOW run state. Two distinct contracts (CheckpointStore vs SnapshotStore), dev-owned pg/redis clients (no peer dep), never-auto-migrated schema(), global defaults via ai.config({defaultCheckpointStore, defaultSnapshotStore}). Triggers: `ai.checkpoint`, `ai.snapshot`, `checkpointStore`, `snapshotStore`, `CheckpointStore`, `SnapshotStore`, `CheckpointRecord`, `checkpoint.pg`, `checkpoint.redis`, `snapshot.pg`, `snapshot.redis`, `store.schema()`, `keepSnapshots`, `defaultCheckpointStore`, `defaultSnapshotStore`, `PgClientLike`, `RedisClientLike`; 'persist orchestrator sessions', 'wire a pg checkpoint store', 'run the store DDL', 'checkpoint vs snapshot'; typical import `import { ai } from "@warlock.js/ai"`. Skip: orchestrator lifecycle — `@warlock.js/ai/run-orchestrator/SKILL.md`; cache-backed snapshot resume / semanticCache store — `@warlock.js/ai/persist-ai-data/SKILL.md`; competing libs `temporal`, `inngest`.
-- [manage-prompts](@warlock.js/ai/manage-prompts/SKILL.md): Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`; 'register a prompt by name', 'resolve a prompt by name@version or tag', 'pin a production tag to a prompt version', 'diff two prompt versions', 'export / import the prompt registry', 'validate a prompt for missing placeholders', 'merge a registered prompt into another'; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.
+- [manage-prompts](@warlock.js/ai/manage-prompts/SKILL.md): Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`, `criteria`; 'register a prompt by name', 'resolve a prompt by name@version or tag', 'pin a production tag to a prompt version', 'diff two prompt versions', 'export / import the prompt registry', 'validate a prompt for missing placeholders', 'validate a prompt against my own criteria / rules', 'merge a registered prompt into another'; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.
 - [observe-ai-flows](@warlock.js/ai/observe-ai-flows/SKILL.md): The core Observer seam — a generic, tool-agnostic observability hook every flow routes its completed ExecutionReport through. Covers the per-flow `observe?: boolean | Observer` option on ai.agent / workflow / supervisor / team, the global registry (registerObserver / getObservers / setObserveAll / isObserveAll / clearObservers), resolveObservers / notifyObservers resolution, the opt-in AgentConfig.captureMessages → AgentReport.messages full-history capture, the onConfigApplied dependency-inversion seam, and that @warlock.js/ai-panoptic is the batteries-included Observer. Triggers: `Observer`, `observe`, `registerObserver`, `getObservers`, `setObserveAll`, `isObserveAll`, `clearObservers`, `resolveObservers`, `notifyObservers`, `FlowObserveOption`, `ExecutionReport`, `captureMessages`, `AgentReport.messages`, `CapturedMessage`, `onConfigApplied`, `observeAll`; 'observe an agent run', 'send finished reports to a collector', 'capture the full message history', 'observe every flow by default', 'wire panoptic / tracing'; typical import `import { ai, registerObserver } from "@warlock.js/ai"`. Skip: structured logging of events — `@warlock.js/ai/log-ai-calls/SKILL.md`; reading the report tree shape (trips / children) — `@warlock.js/ai/run-ai-agent/SKILL.md`; per-call cost / usage rollup — `@warlock.js/ai/handle-ai-errors/SKILL.md`. The batteries-included Observer is the `@warlock.js/ai-panoptic` package.
 - [persist-ai-data](@warlock.js/ai/persist-ai-data/SKILL.md): Persistence delegated to @warlock.js/cache — workflow + supervisor snapshot resume via snapshotStore (4.3.0: now a SnapshotStore from ai.snapshot.*, ⚠ moved off raw CacheDriver), semantic cache + memory via vector-capable CacheDriver, global defaults via ai.config({defaultStore}) + ai.config({defaultSnapshotStore}). Covers drift detection + three recovery paths. Triggers: `ai.config`, `defaultStore`, `defaultSnapshotStore`, `snapshotStore`, `ai.snapshot`, `wf.resume`, `supervisor.resume`, `WorkflowSnapshot`, `SupervisorSnapshot`, `WorkflowDriftError`, `SupervisorDriftError`, `force: true`; 'resume a workflow run', 'configure snapshot store', 'handle signature drift', 'wire pg vector cache'; typical import `import { ai } from "@warlock.js/ai"`. Skip: orchestrator checkpoint/snapshot store factories — `@warlock.js/ai/manage-ai-stores/SKILL.md`; cache driver catalog — `@warlock.js/cache/cache-basics/SKILL.md`; competing libs `temporal`, `inngest`.
 - [pick-ai-provider](@warlock.js/ai/pick-ai-provider/SKILL.md): Choose an AI provider adapter — @warlock.js/ai-openai (shipped, also handles OpenRouter / Azure via baseURL), @warlock.js/ai-anthropic, @warlock.js/ai-bedrock, @warlock.js/ai-google, @warlock.js/ai-ollama — plus cost truth: ModelPricing (per-1M tokens), Usage cost breakdown, the cachedTokens / cacheWriteTokens / reasoningTokens channels, and capability flags. Triggers: `OpenAISDK`, `SDKAdapterContract`, `ModelContract`, `ModelPricing`, `ModelCapabilities`, `sdk.model`, `sdk.embedder`, `capabilities.vision`, `capabilities.structuredOutput`, `capabilities.reasoning`, `capabilities.promptCaching`, `pricing`, `Usage.cost`, `cachedTokens`, `cacheWriteTokens`, `reasoningTokens`, `reasoning.effort`, `cacheControl`, `baseURL`, `provider: "openrouter"`; 'pick a provider', 'openai vs openrouter', 'does this model support vision/reasoning', 'configure pricing', 'how much did reasoning cost', 'prompt cache tokens'; typical import `import { OpenAISDK } from "@warlock.js/ai-openai"`. Skip: agent factory — `@warlock.js/ai/run-ai-agent/SKILL.md`; competing libs raw `openai`, `@anthropic-ai/sdk`, `@aws-sdk/client-bedrock-runtime`.
 - [rag-loaders-and-stores](@warlock.js/ai/rag-loaders-and-stores/SKILL.md): Turn any source into a RagDocument and index it in a production vector store — the document loaders ai.rag.loadText / loadHtml / loadWeb (SSRF-safe via guardedFetch) / loadPdf (lazy pdf-parse peer), plus the swappable stores ai.rag.pgVectorStore({client}) (pgvector + ensureSchema DDL + hnsw/ivfflat index) and ai.rag.cacheVectorStore(driver), both satisfying VectorStoreContract (upsert / query / removeNamespace). Loaders return the exact RagDocument[] that kb.index() consumes — no adapter. Triggers: `ai.rag.loadText`, `ai.rag.loadHtml`, `ai.rag.loadWeb`, `ai.rag.loadPdf`, `loadText`, `loadHtml`, `loadWeb`, `loadPdf`, `ai.rag.pgVectorStore`, `ai.rag.cacheVectorStore`, `pgVectorStore`, `cacheVectorStore`, `VectorStore`, `PgVectorStoreOptions`, `PgVectorStoreInstance`, `ensureSchema`, `schema()`, `RagLoaderResult`, `LoadWebOptions`, `LoadPdfOptions`, `perPage`, `OutboundPolicy`, `guardedFetch`, `hnsw`, `ivfflat`, `pgvector`, `dimensions`, `PgClientLike`, `PDF_PARSE_INSTALL_INSTRUCTIONS`; 'load a website into a knowledge base', 'index a PDF for RAG', 'strip HTML to text for embedding', 'pgvector store for RAG', 'SSRF-safe document fetch', 'one document per PDF page', 'swap the vector store'; typical import `import { ai } from "@warlock.js/ai"`. Skip: the chunk → embed → retrieve → rerank → cite pipeline that consumes these — `@warlock.js/ai/run-ai-rag/SKILL.md`; the raw embedder primitive — `@warlock.js/ai/embed-text/SKILL.md`; cache similarity internals — `@warlock.js/cache/use-cache-similarity/SKILL.md`; competing libs `langchain` loaders, `llamaindex` readers.
 - [record-replay-llm](@warlock.js/ai/record-replay-llm/SKILL.md): Deterministic, offline LLM tests with ai.vcr(model,{path,mode}) — a record/replay decorator over ANY ModelContract that intercepts only complete()/stream(), delegates name/provider/capabilities/pricing to the inner model, and hashes each request against a JSON cassette on disk. Covers the three modes (record / replay / auto), the cassette format, save(), VcrCassetteMissError, streaming round-trip, hashOptions, and composing below fallbackModel. Triggers: `ai.vcr`, `vcr`, `VcrModel`, `VcrOptions`, `VcrMode`, `Cassette`, `CassetteEntry`, `VcrCassetteMissError`, `hashRequest`, `DEFAULT_HASH_OPTIONS`, `mode`, `path`, `hashOptions`, `save`, `cassette`, record, replay, cassette; 'record LLM responses for tests', 'replay model calls offline in CI', 'deterministic agent test without hitting the provider', 'cassette for model calls'; typical import `import { ai } from "@warlock.js/ai"`. Skip: eval scoring + regression gating — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; the Vitest matchers + mockRouter — `@warlock.js/ai/ai-dx-helpers/SKILL.md`; choosing a provider adapter — `@warlock.js/ai/pick-ai-provider/SKILL.md`; competing libs `nock`, `polly.js`.
+- [refine-prompts](@warlock.js/ai/refine-prompts/SKILL.md): Prompt compiler — systemPrompt(...).refined({ model, criteria, store }): humans keep writing human prompt text (dev code or admin-panel textareas); the refined wrapper lazily rewrites it into a model-optimized version via a refiner model on first agent use, pins the result (lockfile posture — recompiled ONLY when the source text, refiner model, criteria, or built-in recipe version change, never silently over time), and serves the pin thereafter. Explicit surfaces: await refined.refine() → the compiled template STRING (placeholders intact — for admin routes, previews, boot warmup, CI; throws PromptRefinementError on failure) and await refined.refinePrompt() → a composable SystemPromptContract with meta.refinedFrom / meta.refinerModel provenance (register it as a next version to unlock ai.prompts.diff review). Placeholder parity is machine-enforced (the exact {{placeholder}} set must survive or the rewrite is rejected after one repair re-ask); the lazy agent path NEVER throws — on refiner failure it warns once and serves the original. store is a structural { get, set } (any @warlock.js/cache CacheDriver); omitted ⇒ the pin lives on the instance for the process lifetime. Triggers: `refined`, `refine`, `refinePrompt`, `materialize`, `RefinedSystemPromptContract`, `RefinedSystemPromptOptions`, `RefinedPromptStoreLike`, `PromptRefineOptions`, `PromptRefinementError`, `refinedFrom`, `refinerModel`, `fresh`, `prompt-refiner`, 'refine a prompt', 'compile a prompt', 'optimize a system prompt', 'rewrite my prompt to be AI-friendly', 'admin-written prompts', 'prompt refinement store'; typical import `import { ai } from "@warlock.js/ai"`. Skip: registry operations (register / resolve / tag / diff / validate) — `@warlock.js/ai/manage-prompts/SKILL.md`; composing prompts from persona + instruction blocks — `@warlock.js/ai/write-system-prompt/SKILL.md`; grading a prompt against rules without rewriting it — validate({ criteria }) in `@warlock.js/ai/manage-prompts/SKILL.md`.
 - [run-ai-agent](@warlock.js/ai/run-ai-agent/SKILL.md): Build agents with ai.agent({...}) — the single-LLM-turn primitive. Covers execute / stream, attachments, structured output, placeholders, events, agent.eval scoring, the judge-safe preset for resilient LLM-as-judge / verdict classifiers (ai.agent.judge / judge: true — lenient JSON parse + repair + never-throw, for Nova-class models), and auto-adapting raw executables in tools:[]. Triggers: `ai.agent`, `ai.agent.judge`, `agent.execute`, `agent.stream`, `agent.eval`, `AgentResult`, `AgentReport`, `AgentToolEntry`, `JudgeConfig`, `JudgeAgentConfig`, `judge`, `repairAttempts`, `streamingToolGuard`, `attachments`, `repair`, `maxTrips`, `sessionId`, `spawnSubAgent`, `SpawnSubAgentSpec`; 'run an agent', 'stream an agent response', 'structured output schema', 'pass image to agent', 'evaluate an agent', 'LLM-as-judge that survives malformed JSON', 'grade with a Nova model without crashing', 'put a supervisor in tools', 'cancel an agent run', 'spawn a one-shot sub-agent with a per-task budget'; typical import `import { ai } from "@warlock.js/ai"`. Skip: tool definition — `@warlock.js/ai/define-ai-tool/SKILL.md`; workflows — `@warlock.js/ai/run-ai-workflow/SKILL.md`; eval matchers / batch / fallback detail — `@warlock.js/ai/ai-dx-helpers/SKILL.md`; competing libs `langchain`, `ai` (Vercel), raw `openai`.
 - [run-ai-rag](@warlock.js/ai/run-ai-rag/SKILL.md): Retrieval-augmented generation with ai.rag({...}) — a chunk → embed → vector-store → retrieve → rerank → cite pipeline that reuses ai.embedder + a @warlock.js/cache CacheDriver. Covers index() / retrieve() / clear() / asTool(), chunking strategies (recursive | markdown | sentence | fixed), Citation / RetrievedChunk provenance, and the opt-in rerankers ai.rag.keywordReranker / ai.rag.llmReranker. Triggers: `ai.rag`, `rag.index`, `rag.retrieve`, `rag.clear`, `rag.asTool`, `RagConfig`, `RagDocument`, `RetrieveOptions`, `RetrieveResult`, `RetrievedChunk`, `Citation`, `ChunkOptions`, `ChunkType`, `ai.rag.keywordReranker`, `ai.rag.llmReranker`, `cacheVectorStore`, `VectorStore`, `topK`, `threshold`, `candidates`; 'build a knowledge base', 'retrieve relevant chunks for a query', 'cite the source of an answer', 'chunk markdown for embedding', 'rerank retrieval results', 'expose retrieval as a tool'; typical import `import { ai } from "@warlock.js/ai"`. Skip: raw single-string embedding — `@warlock.js/ai/embed-text/SKILL.md`; exact + vector LLM-response cache — `@warlock.js/ai/attach-ai-middleware/SKILL.md` (ai.middleware.semanticCache); tool wiring — `@warlock.js/ai/define-ai-tool/SKILL.md`; competing libs `langchain`, `llamaindex`.
 - [run-ai-team](@warlock.js/ai/run-ai-team/SKILL.md): Manager-led multi-agent teams with ai.team({...}) — transparent sugar over ai.supervisor that maps a manager → route/router, members → intents, and a gate → evaluate, returning a REAL SupervisorContract (no new loop, no new contract). Covers the built-in gate strings "quality" (review-then-fix) and "verify" (test-then-fix), a custom gate function, role mapping (roles / gateKey), and the verbatim supervisor pass-throughs (goal / output / state / maxIterations / snapshotStore / on / observe). Triggers: `ai.team`, `TeamConfig`, `TeamGate`, `TeamGateFn`, `TeamMemberValue`, `manager`, `members`, `gate`, `roles`, `gateKey`, `buildQualityGate`, `buildVerifyGate`, `SupervisorContract`, `ReportType`; 'build a team of agents', 'manager that delegates to members', 'review then fix loop', 'test then fix loop', 'quality gate for a multi-agent run', 'report type team'; typical import `import { ai } from "@warlock.js/ai"`. Skip: routing one input to a fixed roster directly — `@warlock.js/ai/run-supervisor/SKILL.md` (team is sugar over it); durable cross-turn sessions — `@warlock.js/ai/run-orchestrator/SKILL.md`; LLM-generated plans — `@warlock.js/ai/run-planner/SKILL.md`; competing libs `crewai`, `autogen`.

package/package.json

@@ -15,14 +15,14 @@
     "@standard-schema/spec": "^1.0.0"
   },
   "peerDependencies": {
-    "@warlock.js/cache": "4.6.0",
-    "@warlock.js/logger": "4.6.0",
+    "@warlock.js/cache": "4.7.0",
+    "@warlock.js/logger": "4.7.0",
     "langfuse": "*",
     "openai": "*",
     "pg": "*",
     "redis": "*"
   },
-  "version": "4.6.0",
+  "version": "4.7.0",
   "main": "./cjs/index.cjs",
   "module": "./esm/index.mjs",
   "types": "./esm/index.d.mts",

package/skills/README.md

@@ -72,6 +72,10 @@ Choose an AI provider adapter — @warlock.js/ai-openai (shipped, also handles O
 
 Deterministic, offline LLM tests with ai.vcr(model,{path,mode}) — a record/replay decorator over ANY ModelContract that intercepts only complete()/stream(), delegates name/provider/capabilities/pricing to the inner model, and hashes each request against a JSON cassette on disk. Covers the three modes (record / replay / auto), the cassette format, save(), VcrCassetteMissError, streaming round-trip, hashOptions, and composing below fallbackModel. Triggers: `ai.vcr`, `vcr`, `VcrModel`, `VcrOptions`, `VcrMode`, `Cassette`, `CassetteEntry`, `VcrCassetteMissError`, `hashRequest`, `DEFAULT_HASH_OPTIONS`, `mode`, `path`, `hashOptions`, `save`, `cassette`, record, replay, cassette; 'record LLM responses for tests', 'replay model calls offline in CI', 'deterministic agent test without hitting the provider', 'cassette for model calls'; typical import `import { ai } from "@warlock.js/ai"`. Skip: eval scoring + regression gating — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; the Vitest matchers + mockRouter — `@warlock.js/ai/ai-dx-helpers/SKILL.md`; choosing a provider adapter — `@warlock.js/ai/pick-ai-provider/SKILL.md`; competing libs `nock`, `polly.js`.
 
+### [`refine-prompts/`](./refine-prompts/SKILL.md)
+
+Prompt compiler — systemPrompt(...).refined({ model, criteria, store }): humans keep writing human prompt text; the refined wrapper lazily rewrites it into a model-optimized version on first agent use, pins the result (lockfile posture — recompiled only when source / model / criteria / recipe change), and serves the pin thereafter. `await refined.refine()` returns the compiled template string (routes / previews / warmup / CI; throws PromptRefinementError); `await refined.refinePrompt()` returns a composable prompt with meta.refinedFrom provenance (register it to diff original vs refined). Placeholder parity machine-enforced; the lazy agent path never throws (falls back to the original, warned once). Load when refining / compiling / optimizing a prompt, wiring a refinement store, or exposing a prompt-refine route. Skip: registry ops — `@warlock.js/ai/manage-prompts/SKILL.md`; the builder itself — `@warlock.js/ai/write-system-prompt/SKILL.md`.
+
 ### [`run-ai-agent/`](./run-ai-agent/SKILL.md)
 
 Build agents with ai.agent({...}) — the single-LLM-turn primitive. Covers execute / stream, attachments, structured output, placeholders, events, AgentResult envelope, streamingToolGuard, the judge-safe preset (ai.agent.judge / judge: true — lenient JSON parse + repair + never-throw for Nova-class LLM-as-judge graders), and ai.spawnSubAgent({...}) (a thin one-shot-agent wrapper with a per-task budget — a general primitive, not planner-specific). Load when calling ai.agent(...), reading AgentResult, wiring options.output / attachments / repair, building a resilient LLM-as-judge, streaming, or spawning a one-shot sub-agent.

package/skills/manage-prompts/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: manage-prompts
-description: 'Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`; ''register a prompt by name'', ''resolve a prompt by name@version or tag'', ''pin a production tag to a prompt version'', ''diff two prompt versions'', ''export / import the prompt registry'', ''validate a prompt for missing placeholders'', ''merge a registered prompt into another''; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.'
+description: 'Unified prompt registry — ai.prompts: one process-wide store of named, versioned systemPrompt(...) builders keyed by name@version. Register by giving a prompt a meta.name (auto-registers), resolve by get(name) / resolve(name, versionOrTag, placeholders) / the inline name@selector form, bulk-register with define(name, versions), pin tags with tag(name, tag, version), compare with diff(name, from, to), round-trip with export() / import(snapshot), and quality-check with a unified validate(target, options) (deterministic missing-placeholder check + optional Nova-safe LLM-as-judge with verdict caching). Compose registered prompts into new ones with systemPrompt().merge(name, { fromVersion }) — provenance recorded in meta.composedFrom. ai.prompt is now a thin FACADE over ai.prompts (BREAKING vs the old standalone registry). Triggers: `ai.prompts`, `ai.prompt`, `PromptsManagerContract`, `PromptsManagerEntry`, `SystemPromptContract`, `SystemPromptMeta`, `SystemPromptMergeOptions`, `PromptsValidateOptions`, `PromptValidationResult`, `PromptValidateTarget`, `PromptTemplateVersion`, `PromptDiff`, `ExportedRegistry`, `defaultPromptsManager`, `prompts()`, `promptKey`, `meta`, `name`, `version`, `composedFrom`, `fromVersion`, `register`, `create`, `get`, `has`, `list`, `versions`, `resolve`, `define`, `tag`, `validate`, `diff`, `export`, `import`, `merge`, `judge`, `judgeCache`, `criteria`; ''register a prompt by name'', ''resolve a prompt by name@version or tag'', ''pin a production tag to a prompt version'', ''diff two prompt versions'', ''export / import the prompt registry'', ''validate a prompt for missing placeholders'', ''validate a prompt against my own criteria / rules'', ''merge a registered prompt into another''; typical import `import { ai } from "@warlock.js/ai"`. Skip: composing a single prompt from persona + instruction blocks (the builder itself) — `@warlock.js/ai/write-system-prompt/SKILL.md`; runtime loadable skill bodies — `@warlock.js/ai/use-runtime-skills/SKILL.md`; eval scoring of agent outputs — `@warlock.js/ai/eval-datasets-and-ci/SKILL.md`; competing libs `langfuse` (direct), `promptfoo`.'
 ---
 
 # `ai.prompts` — the unified prompt registry
@@ -102,6 +102,11 @@ const report = await ai.prompts.validate("support", {
   placeholders: { product: "Warlock" }, // values you intend to supply
   declare: ["language"],                  // extra keys to treat as known
   judge: judgeModel,                      // optional — turns on the LLM-as-judge pass
+  criteria: [                             // optional — YOUR rules, replaces the built-in rubric
+    "Addresses the user by {{name}}",
+    "Never gives medical advice",
+    "Stays under 200 words",
+  ],
 });
 
 report.ok;       // true iff no required placeholder is missing (DETERMINISTIC verdict alone)
@@ -112,6 +117,7 @@ report.issues;   // advisory judge reasons / a degrade note — present only whe
 
 - **Always** runs the deterministic check: every `{{key}}` with no inline default that is neither supplied (`placeholders`), declared (`declare`), nor in the prompt's `meta.required` lands in `missing`; `ok` is `true` iff `missing` is empty.
 - **`judge`** adds a **Nova-safe** LLM-as-judge quality pass — it **never throws** and degrades to an `issues` note (leaving `score` undefined) on failure, so a flaky judge can **never flip `ok`**.
+- **`criteria`** (a string or a list of short rules) grades the prompt against **your own rules** instead of the built-in quality rubric — `score` / `issues` then reflect your criteria (a failed rule is named in `issues`). Only used when `judge` is also set; folded into the `judgeCache` key so different rules re-run. Still advisory — never flips `ok`.
 - **`target`** is a registered name (or `name@selector`), a `SystemPromptContract` instance, or a raw prompt string.
 - **`judgeCache`** (per-call or via the `prompts({ judgeCache })` factory option) memoizes judge verdicts by a content hash of the resolved body + the judge model id — a structural `{ get, set }` subset of `@warlock.js/cache`'s `CacheDriver`, so the cache package stays a strictly **optional** peer.
 
@@ -182,5 +188,6 @@ If you only ever called `ai.prompt({ ... })` and used the returned registry, **n
 ## See also
 
 - [`@warlock.js/ai/write-system-prompt/SKILL.md`](@warlock.js/ai/write-system-prompt/SKILL.md) — the `systemPrompt()` / `persona()` / `instruction()` builder, `.meta()`, and `merge()` this registry stores and composes
+- [`@warlock.js/ai/refine-prompts/SKILL.md`](@warlock.js/ai/refine-prompts/SKILL.md) — `systemPrompt().refined({ model, criteria, store })`, the prompt compiler; register its `refinePrompt()` output as a next version to `diff` original vs refined
 - [`@warlock.js/ai/eval-datasets-and-ci/SKILL.md`](@warlock.js/ai/eval-datasets-and-ci/SKILL.md) — the eval `judge` scorer `validate()`'s LLM pass reuses
 - [`@warlock.js/ai/run-ai-agent/SKILL.md`](@warlock.js/ai/run-ai-agent/SKILL.md) — wiring a resolved prompt into an agent, plus the judge-safe agent preset (`ai.agent.judge`)

package/skills/refine-prompts/SKILL.md

@@ -0,0 +1,91 @@
+---
+name: refine-prompts
+description: 'Prompt compiler — systemPrompt(...).refined({ model, criteria, store }): humans keep writing human prompt text (dev code or admin-panel textareas); the refined wrapper lazily rewrites it into a model-optimized version via a refiner model on first agent use, pins the result (lockfile posture — recompiled ONLY when the source text, refiner model, criteria, or built-in recipe version change, never silently over time), and serves the pin thereafter. Explicit surfaces: await refined.refine() → the compiled template STRING (placeholders intact — for admin routes, previews, boot warmup, CI; throws PromptRefinementError on failure) and await refined.refinePrompt() → a composable SystemPromptContract with meta.refinedFrom / meta.refinerModel provenance (register it as a next version to unlock ai.prompts.diff review). Placeholder parity is machine-enforced (the exact {{placeholder}} set must survive or the rewrite is rejected after one repair re-ask); the lazy agent path NEVER throws — on refiner failure it warns once and serves the original. store is a structural { get, set } (any @warlock.js/cache CacheDriver); omitted ⇒ the pin lives on the instance for the process lifetime. Triggers: `refined`, `refine`, `refinePrompt`, `materialize`, `RefinedSystemPromptContract`, `RefinedSystemPromptOptions`, `RefinedPromptStoreLike`, `PromptRefineOptions`, `PromptRefinementError`, `refinedFrom`, `refinerModel`, `fresh`, `prompt-refiner`, ''refine a prompt'', ''compile a prompt'', ''optimize a system prompt'', ''rewrite my prompt to be AI-friendly'', ''admin-written prompts'', ''prompt refinement store''; typical import `import { ai } from "@warlock.js/ai"`. Skip: registry operations (register / resolve / tag / diff / validate) — `@warlock.js/ai/manage-prompts/SKILL.md`; composing prompts from persona + instruction blocks — `@warlock.js/ai/write-system-prompt/SKILL.md`; grading a prompt against rules without rewriting it — validate({ criteria }) in `@warlock.js/ai/manage-prompts/SKILL.md`.'
+---
+
+# `systemPrompt().refined()` — the prompt compiler
+
+Humans write prompts as human text; models perform better on structured, model-tuned phrasing. `.refined({ model, criteria, store })` turns any `SystemPrompt` into a **lazily-compiled artifact**: the first agent use rewrites the raw source template through the refiner `model`, pins the result, and every later use serves the pin. The human text stays the editing surface forever — the refined text is a derived artifact, like a lockfile.
+
+```ts
+import { ai } from "@warlock.js/ai";
+
+const support = ai
+  .systemPrompt(
+    [ai.persona("You are a friendly assistant."), ai.instruction("Help {{name}} with orders.")],
+    { name: "support" },
+  )
+  .refined({ model: refinerModel, store: myCacheDriver });
+
+// Lazy: compiles on the first run, serves the pin afterwards.
+const agent = ai.agent({ model, systemPrompt: support });
+
+// Explicit: compile now — admin routes, previews, boot warmup, CI.
+const text = await support.refine();          // the compiled template STRING
+const prompt = await support.refinePrompt();  // a composable SystemPromptContract
+```
+
+## The four trust rules
+
+1. **Lockfile posture.** The pin key hashes the recipe version + refiner model + `criteria` + source template — any input change compiles fresh; an unchanged input NEVER recompiles (no TTL, no silent drift). `store` is a **store, not a cache**.
+2. **Prose, never contract.** The exact `{{placeholder}}` set (name **and** `|default`) must survive the rewrite verbatim — checked mechanically; a parity break gets ONE repair re-ask, then the rewrite is rejected. The compiled text is still a **template**: placeholders resolve per call as usual.
+3. **Advisory with fallback.** The lazy agent path never throws: a refiner failure warns once (`[warlock-ai] …`) and serves the ORIGINAL prompt — the human text is always a valid prompt. After **3** failed attempts the lazy path stops retrying for the instance lifetime (no per-run refiner latency from a broken key/provider); the explicit `refine()` / `refinePrompt()` stay live — they **throw** `PromptRefinementError` (`error.reason`: `"model"` / `"parity"` / `"empty"`) and a later success re-arms the pin for everyone.
+4. **Reviewable.** `refine()` exposes the compiled text; `refinePrompt()` makes it a first-class prompt with provenance.
+
+## `refine(options?)` — the explicit string surface
+
+```ts
+const text = await support.refine();                 // store-first; pins on first compile
+const another = await support.refine({ fresh: true }); // skip the pin, new take, re-pins
+```
+
+Expose it via a route for an admin **preview / approve** flow — the admin sees original vs refined, and the call itself warms the pin so the next agent run pays nothing. Also the boot-warmup / CI-compile surface.
+
+## `refinePrompt(options?)` — the composable surface
+
+```ts
+const compiled = await support.refinePrompt();
+
+compiled.blocks;                 // one instruction block = the refined template
+compiled.meta()?.refinedFrom;    // "support@1" (or "anonymous")
+compiled.meta()?.refinerModel;   // "anthropic:claude-sonnet-4-5"
+compiled.meta()?.required;       // carried from the source — contract preserved
+```
+
+It never auto-registers (no `name` — registry versions stay human-intentional). Register it deliberately to unlock the review flow:
+
+```ts
+compiled.meta({ name: "support" });          // registers as support@<next>
+ai.prompts.diff("support", "1", "2");        // original vs refined, block by block
+```
+
+## Options
+
+- **`model`** (required) — the refiner `ModelContract`. The call runs as a one-shot `"prompt-refiner"` agent, so usage/cost surface through the standard report/observer machinery.
+- **`criteria`** — a string or list of rules the rewrite MUST satisfy, on top of the built-in recipe. Same word and shape as `validate({ criteria })`: *validate grades against criteria; refined rewrites against them*. Folded into the pin key — new rules compile fresh.
+- **`store`** — structural `{ get, set }` (`RefinedPromptStoreLike`; any `@warlock.js/cache` `CacheDriver` satisfies it — the cache package stays an optional peer). Share a redis/pg-backed driver so ONE process pays each compilation and the fleet reads the pin. Omitted ⇒ the pin lives on the wrapper instance for the process lifetime. A pinned value that fails the parity check (corrupt / tampered store) is treated as a miss and recompiled.
+
+## What compiles where — the lazy boundary
+
+The lazy compile hook rides the **agent path** (`ai.agent` execute/stream, and everything built on it — supervisors' member agents, planner steps, eval, `spawnSubAgent`, `serve`). Prompts resolved **synchronously at factory time** — `ai.planner({ systemPrompt })` / `ai.router({ systemPrompt })` prefixes, a supervisor's own `systemPrompt` / `goal`, and `ai.prompts.resolve()` — use the ORIGINAL text unless you pre-warm:
+
+```ts
+await refined.refine();                       // warm the pin at boot…
+const planner = ai.planner({ systemPrompt: refined, ... }); // …then factories see it? NO —
+```
+
+Factory-time resolution reads whatever is pinned **at that moment** — so warm BEFORE constructing the factory, or pass `await refined.refinePrompt()` instead (an already-compiled plain prompt).
+
+## Chaining and identity
+
+- `refined.meta()` reads the SOURCE meta — agent reports stamp the source `name@version`, so observability groups by the prompt you authored.
+- `.persona()` / `.instruction()` / `.merge()` / `.meta({...})` derive a NEW source and re-wrap it with the same refinement options — editing a compiled prompt invalidates its pin naturally (new source ⇒ new key).
+- `refined.source` is always the original builder; `refined.resolve(placeholders)` serves the compiled text once pinned, the original before.
+- **Register the source or the `refinePrompt()` output — not the wrapper itself.** The wrapper's `blocks` flip from source to compiled text on materialization, so `ai.prompts.register(wrapper)` would fingerprint whatever is pinned at call time (and a re-register after the flip throws on the content mismatch).
+- `validate()` on the wrapper validates what it currently serves — pair `refined` with `validate({ criteria, judge })` to lint the compiled text, and with `agent.eval` (original vs refined on a dataset) to PROVE the rewrite helps before trusting it.
+
+## See also
+
+- [`@warlock.js/ai/manage-prompts/SKILL.md`](@warlock.js/ai/manage-prompts/SKILL.md) — the registry (`name@version`, tags, `diff`, `validate({ criteria })`) the review flow rides on
+- [`@warlock.js/ai/write-system-prompt/SKILL.md`](@warlock.js/ai/write-system-prompt/SKILL.md) — the `systemPrompt()` builder `.refined()` extends
+- [`@warlock.js/ai/eval-datasets-and-ci/SKILL.md`](@warlock.js/ai/eval-datasets-and-ci/SKILL.md) — measure original vs refined behaviour on a dataset

package/skills/write-system-prompt/SKILL.md

@@ -167,5 +167,6 @@ Three distinct prompts, one common foundation. Base is immutable — safe to sha
 ## See also
 
 - [`@warlock.js/ai/manage-prompts/SKILL.md`](@warlock.js/ai/manage-prompts/SKILL.md) — the `ai.prompts` registry these named prompts auto-register into (resolve / version / tag / diff / export / validate)
+- [`@warlock.js/ai/refine-prompts/SKILL.md`](@warlock.js/ai/refine-prompts/SKILL.md) — `.refined({ model, criteria, store })`, the prompt compiler: lazily rewrite this builder into a model-optimized version, pinned like a lockfile
 - [`@warlock.js/ai/run-ai-agent/SKILL.md`](@warlock.js/ai/run-ai-agent/SKILL.md) — `systemPrompt` on factory + per-call override
 - [`@warlock.js/ai/run-ai-workflow/SKILL.md`](@warlock.js/ai/run-ai-workflow/SKILL.md) — per-step agent references inherit their own system prompt