Self-hosted programming language and toolchain with Python-like syntax, native compilation, integrated testing, and bare-metal support.
This repository also configures its own agents. See what simple tells them →
Agent Claude Code
Use the full bootstrap execution graph. Phase 2 sanity opens three parallel branches: six LLVM/Cranelift subsystem test binaries, Phase 3, and early Phase 4 from Phase 2. Phase 3 sanity opens whole tools/library tests and Phase 4 from Phase 3. Test both Phase 4 cohorts. The selected 80-job code-generation budget must…
Agent Claude Code
Use when: Writing new code, editing existing code, implementing features, fixing bugs. Skills: /coding, /design, /sffi.
Agent Claude Code
Use when: Actively debugging a running session with breakpoints and LSP-enriched inspection. For log-based debugging (env vars, IR export, bug database), use the debug agent instead. Skills: /debug-lsp, /debug Mandatory step order + host-verified commands: .claude/skills/lib/debugladder.md.
Agent Claude Code
Use when: Investigating bugs, analyzing errors, profiling, debugging runtime behavior. Skills: /debug, /mcp-failure-analysis Mandatory step order + host-verified commands: .claude/skills/lib/debugladder.md.
Agent Claude Code
Use when: Writing research docs, design docs, requirement docs, feature specs, guides, reports, updating TODO tracking. Skills: /doc, /todo, /rule, /spipe, /design, /research.
Agent Claude Code
Use when: Finding files, understanding code structure, tracing dependencies, answering "where is X?" or "how does X work?" Skills: /research, /architecture.
Agent Claude Code
Use when: Working on MCP servers, database library, stdlib modules, SFFI wrappers. Skills: /mcp, /database, /stdlib, /sffi.
Agent Claude Code
Use when: Working on ML features, training pipelines, neural networks, dimension checking. Skills: /deeplearning.
Agent Claude Code
This agent requires the following MCP servers (not loaded by default — enable when spawning T32 tasks).
Agent Claude Code
Use when: Writing new tests, fixing failing tests, running test suite, analyzing test results. Skills: /test, /spipe.
Agent Claude Code
This agent requires the following MCP server (not loaded by default — enable when spawning UI design tasks).
Agent Claude Code
Use when: Committing, pushing, syncing, viewing history, managing changes. Skills: /sync.
Agent Claude Code
Use when: RSS is too high or climbing, an OOM/kill, a heap-growth or leak suspicion, or an object that reads back wrong (corrupt aggregate). Related: perf.md (time), debug.md (correctness), build.md.
Agent Claude Code
Use when: Something got slower, a build/lint/test exceeds its budget, a hot path needs a cost model, or a perf gate went red. Related: debug.md (correctness first), mem.md (RSS/heap), build.md.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: