@ancleto/spec 0.6.34 → 0.6.36

package/README.md

@@ -222,6 +222,10 @@ Se guarda en `.ancletorc` → `discovery.exclude` y se preserva en `update`/`upg
 
 La excepción: si el agente detecta un **hueco concreto** (info que el seed debería tener y no tiene, como una paleta nueva sin unidad), puede ofrecer regenerar aunque el `impact` sea `minor` — siempre con tu aprobación. Nunca regenera solo.
 
+### Regeneración parcial
+
+`--check` también devuelve `affectedDocs`: los documentos del seed cuyas áreas cambiaron. Cuando aprobás una regeneración sobre un reporte `minor`, el seed-writer **reescribe solo esos documentos** y deja el resto intacto (un solo pack igual, pero sin reescribir los 8 documentos + dossiers). El mapeo vive en `seed-map.json` dentro del directorio del seed y lo mantiene la skill al generar; sin ese archivo el chequeo degrada a lista vacía y la regeneración es completa.
+
 ---
 
 ## Uso rápido

package/agents/orchestrator.md

@@ -71,7 +71,7 @@ The orchestrator never executes `ancleto discovery --check`. When the technical
 
 For `MISSING` and `PARTIAL`, never generate automatically. Ask for explicit user approval before invoking `@technical-seed-writer`. If approved, pass the complete check report returned by `@technical-discovery` and the action `generate` or `complete`. If declined, continue only when the request can be handled safely without broad repository context; otherwise report the limitation and stop.
 
-For `STALE`, delegate `regenerate` to `@technical-seed-writer` only after explicit user approval. When `impact` is `minor`, keep working from the seed and mention the drift in the summary. A concrete gap found while working — needed info the seed does not cover — may justify offering regeneration even with `minor` impact, always with approval.
+For `STALE`, delegate `regenerate` to `@technical-seed-writer` only after explicit user approval. When `impact` is `minor`, keep working from the seed and mention the drift in the summary. A concrete gap found while working — needed info the seed does not cover — may justify offering regeneration even with `minor` impact, always with approval. When the report lists `affectedDocs`, pass them so an approved regeneration stays scoped to those documents.
 
 `@technical-discovery` owns read-only repository orientation and seed-state inspection. `@technical-seed-writer` owns all seed generation, completion, and regeneration. Never substitute an unavailable seed with a manual repository scan.
 

package/agents/technical-discovery.md

@@ -127,7 +127,7 @@ When you were delegated, do not address the developer as if they had called you:
 - **Answer in Spanish.** Paths, file names, and identifiers stay as they are.
 - Cite the paths that support each claim.
 - State which source you used and whether seed inspection was required.
-- When seed inspection was required, return the resolved state with its `impact`, and the complete unchanged check report if a seed action is needed.
+- When seed inspection was required, return the resolved state with its `impact` and `affectedDocs`, and the complete unchanged check report if a seed action is needed.
 - Declare — when they apply — the stale nodes, any direct code read, and any gap left by the budget.
 - If the seed state prevented the answer, state that no repository sweep was performed and
   include the required `SEED_ACTION_REQUIRED` marker.

package/agents/technical-seed-writer.md

@@ -45,11 +45,11 @@ action was explicitly requested, run `ancleto discovery --check` once and read i
 1. For `READY`, do not rewrite the seed; report that it is already available.
 2. For `MISSING`, `PARTIAL`, or an accepted `STALE` regeneration, execute the skill's single global Repomix command exactly once. It must use the resolved `config.exclude` values and
    must not run `--include`, an uncompressed preliminary pack, or another compressed pack.
-3. Read that pack and immediately write every required seed document in Spanish under
-   `config.outputDir`, including `units/_map.md`. Use the same pack for any optional dossier.
-4. Run `ancleto discovery --check` once to validate the written seed. This is state-only: do
+3. Read that pack and write the seed documents under `config.outputDir`, including `units/_map.md`. When the passed report has `impact: minor` with a non-empty `affectedDocs` list, rewrite ONLY those documents and leave every other seed document untouched. Otherwise (full regeneration), write every required seed document. Use the same pack for any optional dossier.
+4. Write `seed-map.json` in `config.outputDir` mapping every seed document to the source areas it draws evidence from. Areas are first-level directories (or root file names), the same convention the CLI uses. Example: `{ "version": 1, "generatedAt": "...", "docs": { "overview.md": ["src", "package.json"] } }`.
+5. Run `ancleto discovery --check` once to validate the written seed. This is state-only: do
    not generate another pack after it.
-5. Return the result as specified in `## Output`.
+6. Return the result as specified in `## Output`.
 
 If the pack fails, writing fails, or the final state is not usable, report the exact blocker;
 do not claim the seed was generated.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ancleto/spec",
-  "version": "0.6.34",
+  "version": "0.6.36",
   "description": "Orquestador SDD liviano con subagentes optimizados para costo/tokens",
   "type": "module",
   "bin": {

package/skills/ancleto-technical-discovery/references/generation-pipeline.md

@@ -51,6 +51,11 @@ do not split it or create child documents.
 
 Run `ancleto discovery --check` once more and apply `validation-checklist.md`. This is a
 state-only validation; it must not be followed by another Repomix command. For stale nodes,
-regenerate only the affected documents using one new global generation on a later request.
+regenerate only the affected documents using one new global generation on a later request:
+when the check report lists `affectedDocs`, those are the only documents to rewrite.
 If a unit has no dossier, update its row in `units/_map.md` and only the root documents whose
 content changed.
+
+After any generation, write `seed-map.json` in `config.outputDir` mapping every seed document
+to the source areas (first-level directories or root file names) it draws evidence from, so a
+later `--check` can scope the next regeneration.

package/skills/ancleto-technical-discovery/references/validation-checklist.md

@@ -9,4 +9,5 @@
 - [ ] `decisions.md` concentrates rules, risks, and coupling without duplicating the overview.
 - [ ] `inventory.md` declares directories or globs, not file lists.
 - [ ] `unknowns.md` records what could not be verified.
+- [ ] `seed-map.json` exists and maps each seed document to its source areas.
 - [ ] No secret values are present.

package/src/cli/index.js

@@ -1059,8 +1059,24 @@ async function initProject(args) {
     await writeFile(tierFile, tier + '\n')
   }
 
-  const manifest = await writeManifest(projectDir, { azure, discovery, agent, language })
+  // Paridad con install: el tier elegido tambien reescribe los modelos de los
+  // agentes locales. Antes quedaba huerfano (tier declarado, modelos de otro tier).
+  let gratisModelChoice = null
+  if (tier === 'gratis') {
+    const persisted = isKnownGratisModel(existing?.gratisModel) ? existing.gratisModel : null
+    gratisModelChoice = persisted || gratisModel()
+  }
+  const manifest = await writeManifest(projectDir, {
+    azure,
+    discovery,
+    agent,
+    language,
+    ...(gratisModelChoice ? { gratisModel: gratisModelChoice } : {})
+  })
   await copyTemplates(projectDir)
+  if (tier) {
+    await applyTier(join(projectDir, '.opencode', 'agents'), tier, tierModels(tier, TIERS, gratisModelChoice))
+  }
   await scaffoldAspec(projectDir)
   await refreshWorkingContext(projectDir)
   await registerProject(projectDir, { agent, tier: tier || null })
@@ -1207,6 +1223,28 @@ function computeImpact(stored, sources, fileHashes) {
   return { impact, changedAreas: [...changedAreas].sort(), materialReasons, notes }
 }
 
+// seed-map.json lo escribe la skill al generar: cada documento → areas de
+// primer nivel (o archivos raiz) de las que saco evidencia. Misma convencion que areaOf.
+async function readSeedMap(outputDir) {
+  try {
+    const raw = await readFile(join(outputDir, 'seed-map.json'), 'utf8')
+    const parsed = JSON.parse(raw.replace(/^\uFEFF/, ''))
+    if (!parsed || typeof parsed.docs !== 'object' || parsed.docs === null) return null
+    return parsed.docs
+  } catch {
+    return null
+  }
+}
+
+function affectedDocsFor(seedMap, changedAreas) {
+  if (!seedMap || !changedAreas.length) return []
+  const out = new Set()
+  for (const [doc, areas] of Object.entries(seedMap)) {
+    if (Array.isArray(areas) && areas.some((a) => changedAreas.includes(a))) out.add(doc)
+  }
+  return [...out].sort()
+}
+
 function statePath(outputDir) {
   return join(outputDir, '.discovery-state.json')
 }
@@ -1229,7 +1267,7 @@ async function checkDiscovery() {
   const { outputDir } = config
   const tier = readProjectTier(process.cwd())
   const present = await presentDocs(outputDir)
-  let state, action, message, missingDocs, impact = 'none', changedAreas = [], materialReasons = [], notes = []
+  let state, action, message, missingDocs, impact = 'none', changedAreas = [], materialReasons = [], notes = [], affectedDocs = [], seedMapPresent = false
   if (present.length === 0) {
     state = 'MISSING'
     action = 'generate'
@@ -1265,6 +1303,9 @@ async function checkDiscovery() {
       changedAreas = diff.changedAreas
       materialReasons = diff.materialReasons
       notes = diff.notes
+      const seedMap = await readSeedMap(outputDir)
+      seedMapPresent = seedMap !== null
+      affectedDocs = affectedDocsFor(seedMap, changedAreas)
       state = 'STALE'
       if (impact === 'material') {
         action = 'regenerate'
@@ -1286,6 +1327,8 @@ async function checkDiscovery() {
     changedAreas,
     materialReasons,
     notes,
+    affectedDocs,
+    seedMapPresent,
     config: {
       outputDir,
       exclude: config.exclude,