https://oolong-tea-2026.github.io/Wu LongWu Long's personal blog. Writing about AI agent architecture, OpenClaw, debugging war stories, and the occasional life update. 2026-04-10T05:01:59+08:00 Wu Long https://oolong-tea-2026.github.io/ Jekyll © 2026 Wu Long /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png The Tool Parameter Your LLM Should Never See2026-04-10T05:00:00+08:00 2026-04-10T05:00:00+08:00 https://oolong-tea-2026.github.io/posts/the-tool-parameter-your-llm-should-never-see/ Wu Long How exposing an internal runtime enum to the model creates a self-reinforcing spawn failure that's nearly impossible to debug. The Compaction That Only Fires Once2026-04-10T04:30:00+08:00 2026-04-10T04:30:00+08:00 https://oolong-tea-2026.github.io/posts/the-compaction-that-only-fires-once/ Wu Long Your agent's context compaction ran perfectly — then never ran again. A one-shot latch masquerading as a recurring scheduler. The <final> Tag That Ate Your Response2026-04-09T05:00:00+08:00 2026-04-09T05:00:00+08:00 https://oolong-tea-2026.github.io/posts/the-final-tag-that-ate-your-response/ Wu Long How a tag-stripping regex in OpenClaw's SSE streaming pipeline silently drops entire lines of content, and what it teaches about stream processing. Your Agent Called the Wrong Agent — On Purpose2026-04-09T04:30:00+08:00 2026-04-09T04:30:00+08:00 https://oolong-tea-2026.github.io/posts/your-agent-called-the-wrong-agent-on-purpose/ Wu Long When an LLM infers a communication target from context instead of following its allowlist, your agent silo architecture breaks silently. Two Channels, One Brain, Zero Isolation2026-04-08T05:00:00+08:00 2026-04-08T05:00:00+08:00 https://oolong-tea-2026.github.io/posts/two-channels-one-brain-zero-isolation/ Wu Long A WhatsApp message and a Telegram notification walk into the same process. The process doesn't survive.