MIMihir Inamdarinaihive.hashnode.dev·Jul 30 · 28 min readRecall: A Working Content-Addressed Cache for Deep Agents Subagent DispatchMulti-agent frameworks solve context rot by isolating work into subagents: each one gets its own context window, its own tool access, its own billing. The isolation is the point. But it comes with a s00
MIMihir Inamdarinaihive.hashnode.dev·Jul 30 · 24 min readBastion: A Schema-Aware Validator for Incomplete MCP Tool ResponsesAn agent calling a tool through the Model Context Protocol (MCP) (Anthropic, 2024) treats a 200-equivalent JSON-RPC result as ground truth. It has no innate sense of whether the payload it just receiv00
MIMihir Inamdarinaihive.hashnode.dev·Jul 30 · 23 min readVerdict: A Working Risk Classifier for Agentic Software FactoriesCoding agents can now run for hours without a human reading a line of what they produce. This post is the design spec for Verdict, a small library I'm building that gives an agentic coding loop a stru00
MIMihir Inamdarinaihive.hashnode.dev·Jul 30 · 24 min readLedger: A Consumer-Side Contract Guard for MCP Tool-Schema DriftThe Model Context Protocol's 2026-07-28 specification revision added caching hints — ttlMs and cacheScope — to tools/list results, letting a client hold onto a server's declared tool set instead of re00
MIMihir Inamdarinaihive.hashnode.dev·Jul 30 · 37 min readBreaker: A Session-Local Loop Guard for Coding AgentsAn autonomous coding agent chasing a failing test, a build error, or an elusive bug can get stuck in a loop: repeatedly attempting near-identical fixes, re-reading the same file without changing appro00
