Agent Tooling
Local-first tooling that makes agents reliable in real repositories: deterministic retrieval, verifiable edits, MCP servers, and reusable skill packaging.
My bias is pragmatic: agents should do real work without surprising the user.
That means:
- Determinism over magic (same inputs produce the same outputs).
- Auditability over vibes (you can inspect what changed and why).
- Privacy-first defaults (local processing whenever possible).
Retrieval That Scales (Without Embeddings as a Requirement)
llmx
What: Local-first codebase indexer with BM25 search and semantic chunk exports (agents can retrieve only what they need).
Why it matters: Most agent failures in large repos are retrieval failures, not "model" failures. llmx makes retrieval fast, cheap, and debuggable.
Verifiable Editing and Reproducible Builds (Codex Toolchain)
codex-xtreme (includes codex-patcher)
What: An interactive wizard for producing optimized, patched Codex binaries, backed by a verified patch application engine.
Why it matters: This is the "edit loop" for agents made explicit: apply changes reliably, then build/run in a reproducible way.
Primary repo (pin this on GitHub): codex-xtreme
Patch engine (linked inside): codex-patcher
Packaging Domain Expertise for Agents
burn-plugin
What: Claude Code plugin for the Burn deep learning framework, with reusable skills/workflows and evidence-backed references.
Why it matters: "Agent tooling" isn't just code. Packaging knowledge so it is verifiable and reusable is what makes tools scale across teams.
Skill Systems (Available on Request)
cwork
What: A context compiler that assembles "base capabilities + domain primer + project context" into a minimal, task-specific prompt package.
Why it matters: Skill systems are what turn ad-hoc prompting into repeatable workflows (especially across many repos).
MCP Servers (Structured Tool APIs)
pyghidra-lite
What: Token-efficient MCP server that exposes a structured "tool surface" for program analysis workflows (compact output by default, opt-in verbosity).
Registry: Official MCP registry listing: io.github.johnzfitch/pyghidra-lite (v0.1.1, status: active, published 2026-01-29).
Why it matters: Good agents use tools. MCP servers let you build high-signal, low-context interfaces that scale beyond ad-hoc prompts.
LLM Desktop Workflow (Anthropic Ecosystem)
claude-cowork-linux
What: Run the official Claude Desktop app on Linux using compatibility stubs and a bubblewrap sandbox.
Why it matters: Makes Claude a first-class Linux tool without sacrificing isolation.
Note: Unofficial community project; no proprietary Claude code is committed.
Professional UX (No Emojis)
Iconics
What: Semantic icon library (8k+ icons) designed to replace emojis with consistent PNG icons and meaning-based search.
Why it matters: Documentation is a product surface. Consistent visuals improve scannability and trust, especially in technical docs.
Closing thought: My working model: start with algorithms and invariants; use model intelligence to choose among safe, explicit actions.