Blog
Engineering, product, and process write-ups from the ShareValue team.
From five skills to a hardened system
Part two: the full technical story of how five slash commands became a hardened, many-agent development system.
The complete technical evolution since the first paper. Five skills grew shared primitives and a coordination layer; the review went from one reviewer to a four-lens consensus; the pre-push hook grew a WARN-to-BLOCK rule engine that crossed into security and correctness; the danger-zone scan consolidated into one fail-closed script; and the review orchestrator got tool-starved and environment-scrubbed. The through-line: the human's job never expanded — every change added machinery below the human, not autonomy to the agents.
May 23, 2026·Akshaya Murthy·engineeringai-agentsai-opsprocessShipping a Production Stack with AI Coding Agents
A practical workflow for small teams: how we let agents do the repetitive work without ceding judgment.
How we let AI agents handle the repetitive parts of shipping code without ceding judgment on irreversible actions: slash-command verbs, a three-agent consensus review on every push, and a hard-coded danger zone the human always confirms.
May 3, 2026·Akshaya Murthy·engineeringaiworkflowprocess