Skip to content

samihalawa/boe-impact

v0.4.0MIT

Compara artículos del BOE entre dos fechas y separa el cambio textual de sus posibles implicaciones prácticas.

Changelog

0.4.0

  • Add portable local MCP and a Codex marketplace without removing the Claude configuration.
  • Replace letter marks with distinct generated symbols and a small Spanish flag at the border.
  • Synchronize desktop instructions, English base/Spanish listing metadata, review specifications and release notes.
  • Preserve upstream marketplace positioning and reconcile the documentary preview conflict.
  • Public directory approval, local-MCP partner eligibility and model-based review remain separate from package tests.

0.3.1

  • Declare a product-scoped local MCP with bounded JSON-RPC framing, dated demos and deterministic report text.
  • Keep skill execution inside that declared MCP; retain the separate developer CLI.
  • Reject ambient proxy credentials without disabling proxy routing or TLS.
  • Use minimal test subprocess environments; clarify runtime, privacy and unresolved directory-policy claims.

0.3.0 — 2026-10-20

Critique: El diff técnico no mostraba cómodamente el antes y el después. Las respuestas de error del sumario podían parecer búsquedas sin resultados.

Improvement: La nueva función boe_find_articles lleva al usuario a una salida más directamente utilizable; Textos y fechas explícitas, informe de cambios legible y validación del estado del proveedor.

Bounded PLACSP fallback: retain at most 8 MiB, parse only complete prefix entries, identify prefix hashes and unread entries/tombstones, and block continuation before the unread suffix. The general fetch limit remains unchanged.

Shared changes: Spanish-first outcome-led onboarding; offline official-source snapshots explicitly labelled as examples; readable Markdown and print-friendly HTML; doctor; bounded input and overwrite protection; escaped report content; safer XML; explicit source/TLS errors; pagination and parser regression tests.

No adoption uplift, model-based acceptance, hosted MCP service, active CI or official directory approval is claimed.

Local completion: synchronized package/runtime/tool contracts; dataset distributions without a usable URL retain their missing-link warning; all seven workflows receive regression coverage; documentary design with historical-data labels and functional filtering.

0.2.0 — 2026-10-20

Critique: El diff técnico no mostraba cómodamente el antes y el después. Las respuestas de error del sumario podían parecer búsquedas sin resultados.

Improvement: Textos y fechas explícitas, informe de cambios legible y validación del estado del proveedor.

Shared changes: Spanish-first outcome-led onboarding; offline official-source snapshots explicitly labelled as examples; readable Markdown and print-friendly HTML; doctor; bounded input and overwrite protection; escaped report content; safer XML; explicit source/TLS errors; pagination and parser regression tests.

No adoption uplift, model-based acceptance, hosted MCP service, active CI or official directory approval is claimed.

0.1.2

Initial standalone public package. Earlier verification remains historical evidence, not proof of this new release.