developer-tools

You explain the project's validation command. The session ends. Next time, the agent needs the same explanation. Keeping the instruction somewhere is useful only if the next run can find the right instruction again.

This walkthrough saves one lesson from a fictional project, opens it in a new process, and retrieves it through a small knowledge graph. You can inspect the exact saved text and repeat the same checks with the downloadable example.

Your AI coding agent already has a knowledge graph. It is just not yours yet. The model knows GitHub, Stack Overflow, and the public training corpus -- it has no idea that in your project npm should be bun, that RFP is shorthand for acquisition need, or that the email about the Stripe receipt for the Obsidian licence lives in your Fastmail mailbox. This post shows the smallest path to fixing that for both Claude Code and opencode, using Terraphim and the three roles we have published over the last week (Terraphim Engineer, Personal Assistant, System Operator).

Two paths. CLI first.

AI coding agents are making us worse engineers, unless we add discipline back. Here is what we do instead of vibe coding, and how you can do it too in 30 seconds.

AI coding assistants are fast, productive, and occasionally catastrophic. One misplaced rm -rf, one accidental git reset --hard, and hours of uncommitted work vanish.

Jeffrey Emanuel (@Dicklesworthstone) built Destructive Command Guard (dcg): a Rust binary with SIMD-accelerated pattern matching, 49+ security packs, and a fail-open design. It is one of the best tools to come out of the AI agent safety space, and it solved a problem we had been fighting with regex hacks.

This post shows how we integrated dcg with OpenCode using its plugin hook system, so destructive commands are intercepted before they run.

A coding agent uses the wrong flag. You explain the correction, and the command works. The useful question is what happens when another session makes the same mistake.

This walkthrough follows one harmless error from failure to saved correction to actual execution. It uses the Harbour project brief from the memory walkthrough, a released Terraphim agent, and an OpenCode adapter included in the download.

AI coding agents make the same mistakes over and over. We built a learning system that captures failures, stores corrections, and feeds them back into future sessions — turning every error into institutional memory.

Builds on Why Graph Embeddings Matter — the deterministic engine that makes "remember this correction forever" cheap. Apply the pattern in your own project via the Command Rewriting How-to.

You know what is embarrassing? Making the same mistake for the tenth time. Last week, I typed docker-compose up instead of docker compose up. The command failed. I sighed. I corrected it. Three days later? Same thing. Same sigh. Same correction.

Builds on Why Graph Embeddings Matter — the deterministic engine that lets Terraphim store and replay corrections in microseconds. Apply the pattern in your own project via the Command Rewriting How-to.

This is the February 2026 announcement for v1.8.1. For the tested failure → reviewed correction → execution flow, follow the current command-rewriting guide.

We are announcing native hook support for terraphim-agent v1.8.1. This feature captures failed commands from AI agents (Claude Code, Codex, OpenCode) and learns from them, creating a personal knowledge base of mistakes and corrections. No more jq dependency, no more bash wrappers: just terraphim-agent learn hook.

How we use Aho-Corasick automata and knowledge graphs to automatically enforce coding standards across AI coding agents like Claude Code, Cursor, and Aider.

New: see Why Graph Embeddings Matter for the underlying engine that makes these hooks possible — sub-millisecond, deterministic, fully explainable.