peaks-loop 4.0.0-beta → 4.0.0-beta.3

package/dist/src/services/crystallization/crystallization-types.d.ts

@@ -0,0 +1,313 @@
+import { z } from "zod";
+/**
+ * CrystallizationEvent — spec §4.5 / §4.7 / §5.
+ *
+ * M5 scope: the crystallization_event table + the 4-section
+ * evidence_brief projection (spec §4.7 / §10 RL-7). Every
+ * user-facing recommendation is gated on the brief being complete;
+ * the schema below enforces the 4-section shape at parse time.
+ *
+ * Hard rules enforced at the boundary (and re-asserted in the
+ * service layer):
+ *
+ *   1. The 4 brief sections (what_happened / why_it_matters /
+ *      what_learned / what_action) are ALL required. A partial
+ *      brief is a hard parse error (`MISSING_BRIEF_SECTION`),
+ *      and `parseCrystallizationEvent` will throw.
+ *   2. The trigger enum mirrors spec §4.5 exactly:
+ *      user_explicit | llm_suggested | success_default_prompt |
+ *      similar_task_recurrence.
+ *   3. The event's own lifecycle_status is ORTHOGONAL to the
+ *      created/updated loop_release.lifecycle_status — the same
+ *      asset can have many crystallization events over its life.
+ *   4. source_trace_pointers are stable ids (the workflow trace
+ *      id space), not path strings — the column is JSON-array of
+ *      strings.
+ *   5. The brief section guard is implemented as a `.refine(...)`
+ *      on the input schema — the canonical 4 keys must be present
+ *      with non-empty values (the spec mandates 1-2 sentences per
+ *      section; an empty string would be a degenerate brief).
+ *
+ * Out of scope (deferred to later slices):
+ *   - cross-event aggregation (loop crystallized N times).
+ *   - brief rewrite-on-update semantics (M8 dogfood).
+ */
+export declare const CrystallizationTriggerSchema: z.ZodEnum<["user_explicit", "llm_suggested", "success_default_prompt", "similar_task_recurrence"]>;
+export type CrystallizationTrigger = z.infer<typeof CrystallizationTriggerSchema>;
+export declare const CRYSTALLIZATION_TRIGGERS: readonly CrystallizationTrigger[];
+export declare const CrystallizationEventStatusSchema: z.ZodEnum<["candidate", "stable", "retired"]>;
+export type CrystallizationEventStatus = z.infer<typeof CrystallizationEventStatusSchema>;
+/**
+ * EvidenceBrief — the 4-section brief (spec §4.7).
+ *
+ * Hard shape invariant: ALL FOUR KEYS must be present at parse
+ * time, with non-empty NL content. The `.refine(...)` guard is
+ * what makes a brief a "brief" — without it, the projection is
+ * not a brief and the recommendation MUST NOT be rendered
+ * (service layer enforces this; Zod makes it a parse-time
+ * guarantee).
+ *
+ * Field semantics:
+ *   - what_happened   — 1-2 sentence factual account of the run.
+ *   - why_it_matters  — 1-2 sentence explanation of why this is
+ *                       worth promoting.
+ *   - what_learned    — 1-2 sentence learning: failure modes
+ *                       encoded, preferences extracted.
+ *   - what_action     — 1-sentence recommended action with
+ *                       rationale.
+ */
+export declare const EvidenceBriefSchema: z.ZodEffects<z.ZodObject<{
+    what_happened: z.ZodString;
+    why_it_matters: z.ZodString;
+    what_learned: z.ZodString;
+    what_action: z.ZodString;
+}, "strict", z.ZodTypeAny, {
+    what_happened: string;
+    why_it_matters: string;
+    what_learned: string;
+    what_action: string;
+}, {
+    what_happened: string;
+    why_it_matters: string;
+    what_learned: string;
+    what_action: string;
+}>, {
+    what_happened: string;
+    why_it_matters: string;
+    what_learned: string;
+    what_action: string;
+}, {
+    what_happened: string;
+    why_it_matters: string;
+    what_learned: string;
+    what_action: string;
+}>;
+export type EvidenceBrief = z.infer<typeof EvidenceBriefSchema>;
+/**
+ * Helper for downstream code that needs to know whether a parsed
+ * brief satisfies the 4-section rule. Useful for the CLI when a
+ * payload arrived from the LLM and the brief-section guard has
+ * already passed — it returns the section count, always 4 here.
+ */
+export declare function hasAllFourBriefSections(brief: EvidenceBrief): boolean;
+/**
+ * Zod schema for the create payload (CrystallizationEventInput).
+ *
+ * Re-validates the evidence_brief inline; calling code may ALSO call
+ * `EvidenceBriefSchema.parse` separately if it builds the brief out
+ * of band. The CrystallizationEvent input schema wires the same
+ * guard so a brief cannot sneak through via a partial-input path.
+ */
+export declare const CrystallizationEventInputSchema: z.ZodObject<{
+    trigger: z.ZodEnum<["user_explicit", "llm_suggested", "success_default_prompt", "similar_task_recurrence"]>;
+    evidence_brief: z.ZodEffects<z.ZodObject<{
+        what_happened: z.ZodString;
+        why_it_matters: z.ZodString;
+        what_learned: z.ZodString;
+        what_action: z.ZodString;
+    }, "strict", z.ZodTypeAny, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }>, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }>;
+    evidence_bullets: z.ZodDefault<z.ZodArray<z.ZodString, "many">>;
+    source_trace_pointers: z.ZodDefault<z.ZodArray<z.ZodString, "many">>;
+    evaluator_summary: z.ZodDefault<z.ZodString>;
+    user_decision_summary: z.ZodDefault<z.ZodString>;
+    created_loop_release_id: z.ZodOptional<z.ZodString>;
+    updated_loop_release_id: z.ZodOptional<z.ZodString>;
+    created_bee_release_id: z.ZodOptional<z.ZodNumber>;
+    updated_bee_release_id: z.ZodOptional<z.ZodNumber>;
+    lifecycle_status: z.ZodDefault<z.ZodEnum<["candidate", "stable", "retired"]>>;
+}, "strip", z.ZodTypeAny, {
+    lifecycle_status: "candidate" | "stable" | "retired";
+    trigger: "user_explicit" | "llm_suggested" | "success_default_prompt" | "similar_task_recurrence";
+    evidence_brief: {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    };
+    evidence_bullets: string[];
+    source_trace_pointers: string[];
+    evaluator_summary: string;
+    user_decision_summary: string;
+    created_loop_release_id?: string | undefined;
+    updated_loop_release_id?: string | undefined;
+    created_bee_release_id?: number | undefined;
+    updated_bee_release_id?: number | undefined;
+}, {
+    trigger: "user_explicit" | "llm_suggested" | "success_default_prompt" | "similar_task_recurrence";
+    evidence_brief: {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    };
+    lifecycle_status?: "candidate" | "stable" | "retired" | undefined;
+    evidence_bullets?: string[] | undefined;
+    source_trace_pointers?: string[] | undefined;
+    evaluator_summary?: string | undefined;
+    user_decision_summary?: string | undefined;
+    created_loop_release_id?: string | undefined;
+    updated_loop_release_id?: string | undefined;
+    created_bee_release_id?: number | undefined;
+    updated_bee_release_id?: number | undefined;
+}>;
+export type CrystallizationEventInput = z.input<typeof CrystallizationEventInputSchema>;
+/**
+ * Full persisted row schema. Adds `id` (the event id) and stamps
+ * `schema_version` to the fixed literal `peaks.crystallization/1`.
+ *
+ * Brief section guard is INHERITED from `EvidenceBriefSchema.refine`
+ * because `EvidenceBriefSchema` is embedded; the service layer still
+ * re-asserts the guard so a hand-crafted row cannot bypass Zod at
+ * the store boundary.
+ */
+export declare const CrystallizationEventSchema: z.ZodObject<{
+    trigger: z.ZodEnum<["user_explicit", "llm_suggested", "success_default_prompt", "similar_task_recurrence"]>;
+    evidence_brief: z.ZodEffects<z.ZodObject<{
+        what_happened: z.ZodString;
+        why_it_matters: z.ZodString;
+        what_learned: z.ZodString;
+        what_action: z.ZodString;
+    }, "strict", z.ZodTypeAny, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }>, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }, {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    }>;
+    evidence_bullets: z.ZodDefault<z.ZodArray<z.ZodString, "many">>;
+    source_trace_pointers: z.ZodDefault<z.ZodArray<z.ZodString, "many">>;
+    evaluator_summary: z.ZodDefault<z.ZodString>;
+    user_decision_summary: z.ZodDefault<z.ZodString>;
+    created_loop_release_id: z.ZodOptional<z.ZodString>;
+    updated_loop_release_id: z.ZodOptional<z.ZodString>;
+    created_bee_release_id: z.ZodOptional<z.ZodNumber>;
+    updated_bee_release_id: z.ZodOptional<z.ZodNumber>;
+    lifecycle_status: z.ZodDefault<z.ZodEnum<["candidate", "stable", "retired"]>>;
+} & {
+    id: z.ZodString;
+    schema_version: z.ZodDefault<z.ZodLiteral<"peaks.crystallization/1">>;
+    created_at: z.ZodString;
+}, "strip", z.ZodTypeAny, {
+    schema_version: "peaks.crystallization/1";
+    id: string;
+    lifecycle_status: "candidate" | "stable" | "retired";
+    trigger: "user_explicit" | "llm_suggested" | "success_default_prompt" | "similar_task_recurrence";
+    evidence_brief: {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    };
+    evidence_bullets: string[];
+    source_trace_pointers: string[];
+    evaluator_summary: string;
+    user_decision_summary: string;
+    created_at: string;
+    created_loop_release_id?: string | undefined;
+    updated_loop_release_id?: string | undefined;
+    created_bee_release_id?: number | undefined;
+    updated_bee_release_id?: number | undefined;
+}, {
+    id: string;
+    trigger: "user_explicit" | "llm_suggested" | "success_default_prompt" | "similar_task_recurrence";
+    evidence_brief: {
+        what_happened: string;
+        why_it_matters: string;
+        what_learned: string;
+        what_action: string;
+    };
+    created_at: string;
+    schema_version?: "peaks.crystallization/1" | undefined;
+    lifecycle_status?: "candidate" | "stable" | "retired" | undefined;
+    evidence_bullets?: string[] | undefined;
+    source_trace_pointers?: string[] | undefined;
+    evaluator_summary?: string | undefined;
+    user_decision_summary?: string | undefined;
+    created_loop_release_id?: string | undefined;
+    updated_loop_release_id?: string | undefined;
+    created_bee_release_id?: number | undefined;
+    updated_bee_release_id?: number | undefined;
+}>;
+export type CrystallizationEvent = z.infer<typeof CrystallizationEventSchema>;
+/**
+ * Convenience: strict-parse an unknown payload into a
+ * CrystallizationEvent row. Throws ZodError on failure — including
+ * when the brief is missing a section (the `.refine` guard fires).
+ *
+ * Use this at every parse boundary (CLI input, store insertion).
+ */
+export declare function parseCrystallizationEvent(input: unknown): CrystallizationEvent;
+/**
+ * Convenience: strict-parse an unknown payload as a standalone
+ * EvidenceBrief. Throws ZodError on failure.
+ */
+export declare function parseEvidenceBrief(input: unknown): EvidenceBrief;
+/**
+ * Safe-parse twin of parseCrystallizationEvent. Returns a Result-like
+ * envelope so callers (CLI / tests) can render findings without
+ * try/catch noise. The `code` field is `MISSING_BRIEF_SECTION` when
+ * the brief refine guard trips OR when ANY brief-section key is
+ * missing (spec §10 RL-7) — distinguishes brief failures from
+ * generic validation errors.
+ */
+export declare function safeParseCrystallizationEvent(input: unknown): {
+    ok: true;
+    row: CrystallizationEvent;
+} | {
+    ok: false;
+    code?: "MISSING_BRIEF_SECTION";
+    findings: Array<{
+        path: string;
+        message: string;
+    }>;
+};
+/**
+ * Safe-parse twin of parseEvidenceBrief. Same Result-like envelope
+ * shape; same MISSING_BRIEF_SECTION code mapping (covers both
+ * refine-guard failures and per-section missing/empty failures).
+ */
+export declare function safeParseEvidenceBrief(input: unknown): {
+    ok: true;
+    row: EvidenceBrief;
+} | {
+    ok: false;
+    code?: "MISSING_BRIEF_SECTION";
+    findings: Array<{
+        path: string;
+        message: string;
+    }>;
+};

package/dist/src/services/crystallization/crystallization-types.js

@@ -0,0 +1,299 @@
+import { z } from "zod";
+/**
+ * CrystallizationEvent — spec §4.5 / §4.7 / §5.
+ *
+ * M5 scope: the crystallization_event table + the 4-section
+ * evidence_brief projection (spec §4.7 / §10 RL-7). Every
+ * user-facing recommendation is gated on the brief being complete;
+ * the schema below enforces the 4-section shape at parse time.
+ *
+ * Hard rules enforced at the boundary (and re-asserted in the
+ * service layer):
+ *
+ *   1. The 4 brief sections (what_happened / why_it_matters /
+ *      what_learned / what_action) are ALL required. A partial
+ *      brief is a hard parse error (`MISSING_BRIEF_SECTION`),
+ *      and `parseCrystallizationEvent` will throw.
+ *   2. The trigger enum mirrors spec §4.5 exactly:
+ *      user_explicit | llm_suggested | success_default_prompt |
+ *      similar_task_recurrence.
+ *   3. The event's own lifecycle_status is ORTHOGONAL to the
+ *      created/updated loop_release.lifecycle_status — the same
+ *      asset can have many crystallization events over its life.
+ *   4. source_trace_pointers are stable ids (the workflow trace
+ *      id space), not path strings — the column is JSON-array of
+ *      strings.
+ *   5. The brief section guard is implemented as a `.refine(...)`
+ *      on the input schema — the canonical 4 keys must be present
+ *      with non-empty values (the spec mandates 1-2 sentences per
+ *      section; an empty string would be a degenerate brief).
+ *
+ * Out of scope (deferred to later slices):
+ *   - cross-event aggregation (loop crystallized N times).
+ *   - brief rewrite-on-update semantics (M8 dogfood).
+ */
+/* ---------------------------------------------------------------------- */
+/* Trigger — §4.5 / §5.4                                                   */
+/* ---------------------------------------------------------------------- */
+export const CrystallizationTriggerSchema = z.enum([
+    "user_explicit",
+    "llm_suggested",
+    "success_default_prompt",
+    "similar_task_recurrence",
+]);
+export const CRYSTALLIZATION_TRIGGERS = [
+    "user_explicit",
+    "llm_suggested",
+    "success_default_prompt",
+    "similar_task_recurrence",
+];
+/* ---------------------------------------------------------------------- */
+/* Lifecycle status — §5.6 (event-scoped, not asset-scoped)               */
+/* ---------------------------------------------------------------------- */
+export const CrystallizationEventStatusSchema = z.enum([
+    "candidate",
+    "stable",
+    "retired",
+]);
+/* ---------------------------------------------------------------------- */
+/* EvidenceBrief — §4.7 / §10 RL-7. REQUIRED 4-section shape.             */
+/* ---------------------------------------------------------------------- */
+/**
+ * The four brief sections. Spec §4.7 lists them as NL short
+ * sections; the service layer mandates 1-2 sentences (what_action
+ * is a single sentence per the spec). The Zod constraints below
+ * enforce NON-EMPTY strings. The brief-section guard (see
+ * `EvidenceBriefSchema.refine` below) rejects any payload missing
+ * one of the four keys, regardless of section content — it is
+ * the canonical hard gate per spec §10 RL-7.
+ */
+const BriefSectionSchema = z
+    .string()
+    .trim()
+    .min(1, "brief section must be non-empty natural language")
+    .max(4000);
+/**
+ * EvidenceBrief — the 4-section brief (spec §4.7).
+ *
+ * Hard shape invariant: ALL FOUR KEYS must be present at parse
+ * time, with non-empty NL content. The `.refine(...)` guard is
+ * what makes a brief a "brief" — without it, the projection is
+ * not a brief and the recommendation MUST NOT be rendered
+ * (service layer enforces this; Zod makes it a parse-time
+ * guarantee).
+ *
+ * Field semantics:
+ *   - what_happened   — 1-2 sentence factual account of the run.
+ *   - why_it_matters  — 1-2 sentence explanation of why this is
+ *                       worth promoting.
+ *   - what_learned    — 1-2 sentence learning: failure modes
+ *                       encoded, preferences extracted.
+ *   - what_action     — 1-sentence recommended action with
+ *                       rationale.
+ */
+export const EvidenceBriefSchema = z
+    .object({
+    what_happened: BriefSectionSchema,
+    why_it_matters: BriefSectionSchema,
+    what_learned: BriefSectionSchema,
+    what_action: BriefSectionSchema,
+})
+    .strict()
+    .refine((b) => typeof b.what_happened === "string" &&
+    b.what_happened.trim().length > 0 &&
+    typeof b.why_it_matters === "string" &&
+    b.why_it_matters.trim().length > 0 &&
+    typeof b.what_learned === "string" &&
+    b.what_learned.trim().length > 0 &&
+    typeof b.what_action === "string" &&
+    b.what_action.trim().length > 0, {
+    message: "evidence_brief must contain all 4 sections (what_happened, why_it_matters, what_learned, what_action) with non-empty content (spec §4.7 / RL-7)",
+    path: [],
+});
+/**
+ * Helper for downstream code that needs to know whether a parsed
+ * brief satisfies the 4-section rule. Useful for the CLI when a
+ * payload arrived from the LLM and the brief-section guard has
+ * already passed — it returns the section count, always 4 here.
+ */
+export function hasAllFourBriefSections(brief) {
+    return (typeof brief.what_happened === "string" &&
+        brief.what_happened.trim().length > 0 &&
+        typeof brief.why_it_matters === "string" &&
+        brief.why_it_matters.trim().length > 0 &&
+        typeof brief.what_learned === "string" &&
+        brief.what_learned.trim().length > 0 &&
+        typeof brief.what_action === "string" &&
+        brief.what_action.trim().length > 0);
+}
+/* ---------------------------------------------------------------------- */
+/* CrystallizationEvent — §4.5                                            */
+/* ---------------------------------------------------------------------- */
+/**
+ * Structurally referenced columns: pointers to optional loop_release /
+ * bee_release rows. All four are optional — a crystallization event
+ * may exist for "trace only" / "discard" choices that touch no durable
+ * asset. Set is governed by the service layer's pre-run gate.
+ */
+const OptionalLoopId = z
+    .string()
+    .min(1)
+    .max(64)
+    .regex(/^[a-z][a-z0-9]*(-[a-z0-9]+)*$/, {
+    message: "loop_release_id must be kebab-case starting with a lowercase letter",
+})
+    .optional();
+const OptionalBeeId = z
+    .number()
+    .int()
+    .positive()
+    .max(2 ** 31 - 1, "bee_release_id out of 32-bit range")
+    .optional();
+/**
+ * Zod schema for the create payload (CrystallizationEventInput).
+ *
+ * Re-validates the evidence_brief inline; calling code may ALSO call
+ * `EvidenceBriefSchema.parse` separately if it builds the brief out
+ * of band. The CrystallizationEvent input schema wires the same
+ * guard so a brief cannot sneak through via a partial-input path.
+ */
+export const CrystallizationEventInputSchema = z.object({
+    trigger: CrystallizationTriggerSchema,
+    evidence_brief: EvidenceBriefSchema,
+    evidence_bullets: z
+        .array(z.string().trim().min(1).max(1000))
+        .default([]),
+    source_trace_pointers: z
+        .array(z.string().trim().min(1).max(256))
+        .default([]),
+    evaluator_summary: z
+        .string()
+        .trim()
+        .max(4000)
+        .default(""),
+    user_decision_summary: z
+        .string()
+        .trim()
+        .max(4000)
+        .default(""),
+    created_loop_release_id: OptionalLoopId,
+    updated_loop_release_id: OptionalLoopId,
+    created_bee_release_id: OptionalBeeId,
+    updated_bee_release_id: OptionalBeeId,
+    lifecycle_status: CrystallizationEventStatusSchema.default("candidate"),
+});
+/**
+ * Full persisted row schema. Adds `id` (the event id) and stamps
+ * `schema_version` to the fixed literal `peaks.crystallization/1`.
+ *
+ * Brief section guard is INHERITED from `EvidenceBriefSchema.refine`
+ * because `EvidenceBriefSchema` is embedded; the service layer still
+ * re-asserts the guard so a hand-crafted row cannot bypass Zod at
+ * the store boundary.
+ */
+export const CrystallizationEventSchema = CrystallizationEventInputSchema.extend({
+    id: z
+        .string()
+        .min(1)
+        .max(128)
+        .regex(/^crys-[0-9a-f]{8,}$/, {
+        message: "id must start with 'crys-' followed by a hex suffix (spec §4.5)",
+    }),
+    schema_version: z
+        .literal("peaks.crystallization/1")
+        .default("peaks.crystallization/1"),
+    created_at: z.string().datetime(),
+});
+/**
+ * Convenience: strict-parse an unknown payload into a
+ * CrystallizationEvent row. Throws ZodError on failure — including
+ * when the brief is missing a section (the `.refine` guard fires).
+ *
+ * Use this at every parse boundary (CLI input, store insertion).
+ */
+export function parseCrystallizationEvent(input) {
+    return CrystallizationEventSchema.parse(input);
+}
+/**
+ * Convenience: strict-parse an unknown payload as a standalone
+ * EvidenceBrief. Throws ZodError on failure.
+ */
+export function parseEvidenceBrief(input) {
+    return EvidenceBriefSchema.parse(input);
+}
+/**
+ * Safe-parse twin of parseCrystallizationEvent. Returns a Result-like
+ * envelope so callers (CLI / tests) can render findings without
+ * try/catch noise. The `code` field is `MISSING_BRIEF_SECTION` when
+ * the brief refine guard trips OR when ANY brief-section key is
+ * missing (spec §10 RL-7) — distinguishes brief failures from
+ * generic validation errors.
+ */
+export function safeParseCrystallizationEvent(input) {
+    const r = CrystallizationEventSchema.safeParse(input);
+    if (r.success)
+        return { ok: true, row: r.data };
+    const findings = r.error.issues.map((i) => ({
+        path: i.path.join("."),
+        message: i.message,
+    }));
+    const isBrief = isBriefSectionFailure(r.error.issues);
+    return isBrief
+        ? { ok: false, code: "MISSING_BRIEF_SECTION", findings }
+        : { ok: false, findings };
+}
+/**
+ * Safe-parse twin of parseEvidenceBrief. Same Result-like envelope
+ * shape; same MISSING_BRIEF_SECTION code mapping (covers both
+ * refine-guard failures and per-section missing/empty failures).
+ */
+export function safeParseEvidenceBrief(input) {
+    const r = EvidenceBriefSchema.safeParse(input);
+    if (r.success)
+        return { ok: true, row: r.data };
+    const findings = r.error.issues.map((i) => ({
+        path: i.path.join("."),
+        message: i.message,
+    }));
+    const isBrief = isBriefSectionFailure(r.error.issues);
+    return isBrief
+        ? { ok: false, code: "MISSING_BRIEF_SECTION", findings }
+        : { ok: false, findings };
+}
+/**
+ * Decide whether a ZodIssue[] is a "brief section" failure: every
+ * issue points at one of the four required keys (the section is
+ * missing, or empty after trim), OR the explicit refine-guard issue
+ * is present. Used by both `safeParseCrystallizationEvent` and
+ * `safeParseEvidenceBrief` so callers see a single
+ * `MISSING_BRIEF_SECTION` code regardless of which Zod boundary
+ * caught the failure.
+ */
+function isBriefSectionFailure(issues) {
+    if (issues.length === 0)
+        return false;
+    // Explicit refine guard (EvidenceBriefSchema.refine message).
+    if (issues.some((i) => i.message.includes("evidence_brief must contain all 4 sections"))) {
+        return true;
+    }
+    // Per-section Zod failures (min(1) on each section key, or
+    // too_small, or invalid_type). Each issue's path will be either
+    // the section key (`what_happened`) OR the parent `evidence_brief`
+    // — Zod reports missing-key issues against the parent object.
+    return issues.every((i) => {
+        const path = i.path.join(".");
+        if (path === "evidence_brief")
+            return true;
+        if (i.path.length >= 2 && i.path[0] === "evidence_brief") {
+            const key = i.path[i.path.length - 1];
+            return (key === "what_happened" ||
+                key === "why_it_matters" ||
+                key === "what_learned" ||
+                key === "what_action");
+        }
+        return (path === "what_happened" ||
+            path === "why_it_matters" ||
+            path === "what_learned" ||
+            path === "what_action");
+    });
+}