Obsidian

Open-source compiler for AI agent communication

Developer Ibrahim Kobeissy published /calibrate-comms, an open-source skill implemented inside an Obsidian vault that captures and persists user preferences for how AI agents should write. The tool calibrates nine operational dials (Density, Sequence, Modality, Abstraction, Tradeoff, Detail, Jargon, Tone, Context-giving) via a Quick mode (forced-choice proxies) or Deep mode (fetching open-access instruments when available). Pairwise sample reactions override priors, and a deterministic band-to-rule compiler emits executable directives written into a delimited COMMS-PROFILE block (stored in CLAUDE.md) so multiple agents (notably Claude Code and Codex) can share one source of truth. The author emphasises privacy by keeping populated profiles in a private vault while shipping the public template and an empty placeholder on GitHub. Publication date: 2026-06-26.

Synthadoc v0.3.0 Turns Video and Web into Wiki

Synthadoc v0.3.0 expands the open-source knowledge-engineering tool to ingest YouTube video captions and the live web, converting diverse inputs into structured, cross-referenced Markdown wiki pages. The release preserves [MM:SS] timestamps from video captions, generates LLM-written executive summaries, and auto-creates wikilinks to related pages. A new "web search fan-out" decomposes queries into sub-questions, synthesizes top results into multiple pages, and builds cross-references. The unified ingest pipeline now supports PDFs, Office files, images (via vision LLM), web pages, YouTube, search results, and presentations. The project is AGPL-3.0 on GitHub, includes an Obsidian plugin, and defaults to Gemini Flash as the model provider (with alternatives like Claude Code and Opencode). Release date: 2026-05-04.

10-Agent AI Product Team in Claude Code

A developer describes building a 10-agent AI product team using Claude Code's Agent Teams feature to orchestrate product development stages (ideation through go-to-market). Each agent is defined as a markdown file in a .claude/agents folder and runs in its own context; agents communicate directly and a lead orchestrator ('Athina') enforces stage gates and runs 'Grill Me' challenge sessions. The author migrated from an OpenClaw setup to Claude Code to reduce infrastructure friction and token costs, splitting agents across Opus 4.6 (open-ended reasoning) and Sonnet 4.6 (procedural checklist work). The workflow uses the Superpowers plugin to enforce TDD, Playwright for E2E QA, and a Codex (GPT) adversarial review step to provide cross-model code review. The post highlights cost, portability, and design-alternatives before commitment.