Claude Intelligence Layer — token-efficient context runtime for Claude Code. Persistent memory (SQLite FTS5), structured workflows (/develop /review /commit /wrap-up), output compression, and MCP integration.
This repository also configures its own agents. See what cil tells them →
Command Claude Code
Detect project type (Node/Python/Rust/Go/mixed) and propose CLAUDE.md + settings.json additions for lint, typecheck, and test commands. Pure inspection — applies nothing without explicit user approval.
Command Claude Code
Generate a conventional commit message for staged changes (feat/fix/refactor/...) and run git commit. Refuses if sensitive files are staged.
Command Claude Code
Structured development cycle — Research → Feasibility → Plan → Implement → Verify with persistent memory and feedback replay.
Command Claude Code
Fast session handover — capture current state and write a resume command for the next session. Use when context is filling up or you need to switch tasks.
Command Claude Code
Persist a single insight, decision, or constraint to long-term memory with auto-classified category and tags.
Command Claude Code
Structured codebase orientation — understand an unfamiliar project before making changes. Maps architecture, conventions, entry points, and gotchas.
Command Claude Code
Project-local release pipeline — bump version, regenerate CHANGELOG from commits since last tag, update README docs for new features (with concrete examples + argument explanations), commit + tag + push, open a PR if on a non-main branch, then hand off a manual npm publish (no token configured yet).
Command Claude Code
Full-text search persistent memory (BM25) and present results grouped by category — decisions, constraints, learnings, tasks.
Command Claude Code
Review the working diff (unstaged + staged) against a correctness, regressions, security, performance, and style checklist.
Command Claude Code
Capture testable acceptance criteria for a feature before implementation. BA hat — ambiguities surfaced, scope bounded, success defined.
Command Claude Code
Generate a test scaffold for a file or function following the testing skill protocol — boundary, happy path, edge cases, error cases. Does not run; produces the file.
Command Claude Code
End-of-session protocol — capture decisions, constraints, learnings, and next-session task; create a session snapshot before context is lost.
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: