Posts 264 entriesPage 8 of 27

Posts

Technical Deep-Dive

Ninety Skills, Three Packs, One Day

Building three complete 30-skill SaaS integration packs — Sentry, Notion, and Supabase — in a single day. Plus cad-dxf-agent v0.11.0 with architectural drift detection as a CI gate.

Read
Technical Deep-Dive

Nuclear Option Day: Validator Rewrite, 63 New Packs, 414 Plugins

Rewrote the universal validator from scratch to match Anthropic's 14-field spec, generated 63 SaaS packs, scored every skill in the catalog, and eliminated all D/F grades. 40+ commits in one day.

Read
Technical Deep-Dive

58 E2E Tests, a Slack Channel Launch, and the Auth Injection That Made It Work

How REST API auth plus IndexedDB injection unlocked 58 production E2E tests for cad-dxf-agent, while claude-code-slack-channel went from zero to v0.1.0 with CI and upstream submission in one day.

Read
Technical Deep-Dive

Governed Knowledge: Two Major Releases, a Freshness Daemon, and Export Gating

qmd-team-intent-kb ships v0.2.0 and v0.3.0 in one day — freshness-aware search with exponential decay, an edge daemon that auto-deprecates stale memories, security hardening, and export gating. Plus claude-code-plugins v4.20.0.

Read
Development Journey

Knowledge Base from Zero to API: Lifecycle State Machines and Policy Engines

Building a team knowledge base with lifecycle state machines, policy engines, and git-exported articles — from empty repo to working API in one day.

Read
Development Journey

Content Quality War: 7-Check Audit Across 340 Plugins

A 7-category audit script found boilerplate openings, empty shells, reference stubs, and duplicate content across 340 plugins. Every category got its own fix batch.

Read
AI Engineering

Meta-Agent Orchestration and the UX of Removing Features

The oss-agent-lab ships its meta-agent capstone — one agent to orchestrate nine specialists. Meanwhile, the plugin marketplace learns that removing features can be progress.

Read
AI Engineering

Building a Meta-Agent System From Scratch in One Day

From first commit to a 9-specialist multi-agent system with capability scoring, session memory, and a meta-agent orchestrator. All in one day.

Read
Development Journey

Terminal-Bold Redesign and the 0% Recall Bug

Phase 2 of the marketplace facelift brings monospace headings and sharper contrast. Plus: why the eval script reported 0% recall for skills that were clearly working.

Read
Technical Deep-Dive

OKLCH: Why Your CSS Color System Is Lying to You

HSL colors look mathematically equal but perceptually wrong. OKLCH fixes this. How a full marketplace facelift exposed the problem.

Read