The latest articles on Forem by Matthew (@mattbaconz). https://forem.com/mattbaconz https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3886323%2F25b6ddfb-79b0-48c8-a854-5440ad226722.png https://forem.com/mattbaconz en Matthew Sat, 18 Apr 2026 16:49:12 +0000 https://forem.com/mattbaconz/signal-shrinking-agent-skills-so-context-stays-for-code-40p1 https://forem.com/mattbaconz/signal-shrinking-agent-skills-so-context-stays-for-code-40p1 <p><strong>SIGNAL</strong> is a small open-source <strong>agent skill bundle</strong> for people who hit <strong>context limits</strong>: tiered modes, symbolic shorthand, checkpoints, a disk-backed <code>.signal_state.md</code>, and separate skills for <strong>diff/search summaries</strong> and <strong>git</strong> (commit / push / PR) plus <strong>structured review</strong>. Each skill ships as readable <code>*.md</code> and a tight <code>*.min.md</code> so you can load less instruction surface when you want.</p> <h3> Install </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>npx skills add mattbaconz/signal </code></pre> </div> <h2> Repo: github.com/mattbaconz/signal </h2> <p><strong>Benchmarks (honest caveat)</strong><br> The repo includes scripted fixtures and heuristic token estimates (ceil(chars/4)). That’s useful for comparing shapes, not for quoting exact billed tokens from a provider.</p> <p><strong>What I’d like from readers</strong><br> If you’ve tried skills packs or fat AGENTS.md files: does the canonical + minified split match how you’d maintain this, or is it too much friction? I’m open to blunt feedback.</p> ai opensource productivity claude