Build and query architecture-aware dependency graphs and belief maps for multi-language codebases.
npx skills add https://github.com/diskd-ai/codespaces --skill codespaces-diskd-ai
Please help me install this Agent Skill. Skill: codespaces Source: https://github.com/diskd-ai/codespaces Command: npx skills add https://github.com/diskd-ai/codespaces --skill codespaces-diskd-ai
.belief_map.sexp graph for Python, TypeScript/TSX, Rust, C#, Java, Go, Ruby/Rails, and Free Pascal/Lazarus projects, with optional LSP-enriched call edges. - Dependency and impact analysis: Query dependencies, reverse dependencies (blast radius), call chains, callers, architecture-layer violations, and the minimal boundary files for a change. - Infrastructure topology: Extract service relationships from Kustomize, Helm, and Terraform files and merge them into the graph. - Use Case: Before fixing a bug in a monorepo, run quick "file upload" to locate the owning module, see its dependencies and blast radius, and read only the 2-3 boundary files instead of scanning whole directories. ## Quick Start Ask your agent to use the Codespaces skill to build the belief map for your repository, then find where a feature is implemented, analyze that module, and list the boundary files needed for the change.High-intent search queries and answers about installing and using this skill.
Build the belief map with build_belief_map.py, then run `belief_search.py rdeps <module-id> 2` to see reverse dependencies (blast radius) or `boundary <module-id>` to list the minimal files to read. The `analyze` command combines dependencies, reverse dependencies, layers, and violations in one output.
It indexes Python, TypeScript, TSX, Rust, C#, Java, Go, Ruby, Ruby on Rails, Free Pascal, and Lazarus. Python uses the standard library AST; other languages use tree-sitter parsers installed per language only when detected in the target repository.
Yes. It builds a project-local constant index following Zeitwerk naming rules, resolving concerns, associations, callbacks, jobs, mailers, and impacted specs statically. Ambiguous or polymorphic associations remain unlinked rather than producing guessed edges.
The builder loads tree-sitter grammars only for languages detected in your repository and stops before publishing if one is missing or has the wrong version. It prints the exact install command, such as `python3 -m pip install -r requirements/typescript.txt`; run it and rebuild.
Yes. Use `find_callchain <source-fn> <target-fn> [depth]` to trace a call path, `find_callers` for incoming calls, and `find_calls` for outgoing calls. For precise call-site data, rebuild the map with the `--lsp` flag when the relevant language server is installed.
The default map comes from deterministic static parsing, so generated imports, runtime dependency injection, reflection, and framework behavior not explicit in source may be missed. Optional LSP enrichment adds call hierarchy detail, but dynamic behavior should be verified against the running application.
xberg-ioManage tree-sitter parser cache with prefetch, list, inspect, and clean commands.
kagentiCheck code changes for architecture conformance across modules.
nguyenphpRun linters, type checkers, and security audits across Node.js and Python projects.
massgenLocate text and structural code patterns across large codebases using ripgrep and ast-grep.