ACP Build — Edge Cases and Pre-Build Checklist
ACP's pre-build checklist identifies critical blockers including patent-before-publication timing (especially for EU/Japan where no grace period exists), npm scope availability for package naming, and legal conflicts between Apache 2.0's patent grant and the open-core business model. Edge cases requiring resolution span technical concerns—content_hash stability across platforms, YAML parser type-coercion breaking hashes, enrichment idempotency rules, filesystem adapter scaling limits—and operational risks like API key exposure in config files, uncontrolled LLM batch-enrichment costs, and the threat of MCP governance defining a competing content schema that could make ACP redundant.
ACP Build Plan — SDK, Enrichment, and Spec Website
ACP (Atomic Content Protocol) is a planned system comprising three workstreams: an open-source SDK monorepo with core schema validation, LLM-powered enrichment pipelines, and MCP server packages; enrichment tools integrated into Stacklist MCP that use a single LLM call to generate tags, summaries, classifications, and key entities for content objects; and a dedicated spec website. The SDK architecture extracts and generalizes existing Stacklist MCP code into reusable packages (@acp/core, @acp/enrichment, @acp/mcp) with a CLI, supporting multiple AI providers, filesystem-based storage, tiered enrichment depth levels, and 15 MCP tools for creating, enriching, searching, and managing Atomic Content Objects.
Phase 6c — Public ACO Registry + Wiki Flows (new Linear ticket)
Phase 6c turns the hosted ACP MCP server at `mcp.atomiccontentprotocol.org` into a public, content-addressable ACO registry with lookup, publish, find-related, and stats tools, while Stacklist MCP dual-writes enrichments to both its private Supabase database and the public registry with a read-through fallback chain. It also introduces DWIM behavior for `enrich_url` on stacklist.com URLs, where card/stack owners get in-place enrichment updates while non-owners contribute enrichments only to the public registry without modifying the owner's private data—establishing a wiki-like collaborative dynamic.
Phase 6b — Stacks as ACOs (new Linear ticket)
Phase 6b promotes stacks to first-class ACOs by adding full enrichment metadata columns (classification, key_entities, token_counts, relationships, content_hash, etc.) to the stacks table, along with a new `enrich_stack` tool that rolls up child card data, checks cache, and populates ACO frontmatter. It also updates `get_stack_contents` and `get_card_content` to return complete ACO fields in their responses and extends `capsule_from_content` with a `save_as: 'stack'` option for creating enriched stacks directly.
Phase 6a — Stacklist MCP Enrichment Core (LAB-212)
Phase 6a of the Stacklist MCP project ships five enrichment tools (`enrich_content`, `enrich_url`, `enrich_batch`, `capsule_from_content`, `capsule_from_url`) that transform raw text or external URLs into Atomic Content Object (ACO) frontmatter, optionally saving capsules to a user's private Stacklist vault with a content-hash-based cache to skip redundant LLM calls. The spec details the full architecture flow, file map across `stacklist-mcp` and `acp-sdk` repos, eight ordered implementation steps gated by validation, locked design decisions (e.g., `sourceType='manual'`, batch max of 25, cache writes only on save), and a verification plan covering unit tests with mocked dependencies and integration tests against a real Supabase database.