code-memory-skillv1.0.0

A new engineer joins a team maintaining a large, sparsely documented Git repository and needs to quickly understand its architecture. They use code-memory to run topic_discovery searches for concepts like 'authentication middleware' and file_structure queries to map modules before touching any code. This shortens ramp-up time from days to hours while keeping direct file reads and tests as the final source of truth.

A developer planning a risky refactor of a service needs reliable definition and references lookups without running slow language servers across a monorepo. Using search_code with definition and references types, they trace UserService-like symbols and their call sites, then validate findings with rg and typechecker output. This reduces missed call sites and false positives during the refactor.

An SRE investigates a timeout regression introduced weeks ago and needs commit-level context behind a specific file region. They use search_history with commits and file_history query types plus blame data to identify the offending change and its author rationale. This accelerates root-cause analysis while avoiding guesswork from semantic search alone.

A technical writer or API maintainer needs to find how deployment, architecture, or configuration is described across READMEs and docstrings. They run search_docs queries with topic-oriented keywords and top_k tuning to assemble consistent documentation. Retrieved snippets are then verified against source files before publication.

A platform team prepares a major release and wants to prune unused functions, classes, and methods to reduce maintenance surface. They run find_dead_code with min_confidence thresholds and manual verification, cross-checking with tests and runtime logs before deletion. This reduces bundle size and cognitive load without blindly deleting reachable code.

code-memory is distributed free under an open-source license, with paid support, SLAs, and integration consulting for enterprises that need audit trails and security reviews. The upstream GitHub repository drives adoption while revenue comes from support contracts and custom deployments.

A managed SaaS wraps the local MCP server with shared team indexes, per-repo embeddings, and access controls, removing the need for each developer to download roughly 1 GB of model weights. Teams pay per seat or per indexed repository, with data residency options for regulated industries.

The skill is packaged as a premium extension or MCP marketplace listing for AI coding assistants such as Claude, Cursor, or custom agents. Value-added features include curated prompts, dead-code reports, and Git blame summaries, sold through app stores or bundled into enterprise AI coding licenses.