v1.1.0 — Broader production coverage

production

• Broaden audit routing across MCP servers/clients/hosts, agents, self-hosted and disconnected deployments, offline synchronization, real-time/event systems, games/spatial interfaces and embedded/device software.
• Add seven independently applicable parts: MCP, agent execution, deployment and operability, offline/synchronization, real-time/event systems, device/embedded software and software supply chain. Add data-pipeline, model-lifecycle, SDK/CLI, non-web accessibility and platform-compatibility checks: 39 parts and 753 checks, computed from the shipped registry.
• Keep existing check IDs, schema version 1.0, scoring weights and decision rules; extend schema part/surface identifiers. Existing scoped payloads retain their results; full audits must assess the newly applicable parts or explain their exclusion.
• Add dated primary-source references, version-aware protocol guidance and an illustrative MCP-part payload. Distinguish engineering coverage from specialist certification and identify the retained 2025 LLM baseline rather than implying a complete newer-edition mapping.

Distribution

• Link the MCPRush publisher home and individual listings for BrainLLM and both skills; identify the repository commands as the complete-package install path. Both skill package versions move together; blueprint behavior is unchanged.

v1.0.1 — Structure-independent

Both skills stop assuming anything about how a project is laid out.

blueprint no longer dictates a blueprint/ folder at the repository root. It finds an existing blueprint by its index or its documents wherever the project keeps them, and when there is none it asks the owner where it should live, recommending the project's own convention for agent context. What it holds to instead is completeness: the index lists all seven documents with a status, every listed file exists, and nothing else in the folder claims to be intent.

production no longer looks for a blueprint/ folder or any particular file. Its scope phase reconnoitres the project for stated intent (blueprints, PRDs, specs, agent instruction files, READMEs, docs, pricing pages, store listings), uses the strongest statement of requirements it finds as the functional standard, and records what it used in run.assumptions. The brand, cost, functional and analytics references follow suit.

V1.0.0

First public release.

blueprint

• Establishes a project's foundation together with its owner in seven documents: vision.md, market.md, business.md, product.md, brand.md, stack.md and plan.md.
• Each document goes through a per-document cycle: Discuss → Research and compute → Draft → Review → Approve, with owner sign-off.
• blueprint/README.md indexes the documents, workbook.md records progress, and research/ holds dated evidence and a runnable economic model, so the work can resume in any session.
• The blueprint sets intent and leaves design, architecture, schemas, APIs and copy to the builder.

production

• An autonomous production readiness audit: 32 parts and 604 checks, organised by surface (web, API, webhooks, jobs, mobile, desktop, extension, CLI/SDK, messaging, docs).
• A structured JSON payload (assets/payload.schema.json) and a Markdown report that rank parts and surfaces by risk and give a GO, GO WITH CONDITIONS, HOLD or NO-GO decision per surface and overall.
• scripts/readiness.py does the validation, scoring, ranking and gating (Python 3.8+, standard library only).
• Covers OWASP Top 10:2025, the OWASP LLM Top 10, WCAG 2.2 AA, and the major privacy, accessibility, consumer and tax regimes.

Install

npx skills add miisodev/skills

If these skills help you, consider sponsoring or donating via PayPal.