2026-10-24T03:26:54Z https://nostr.now ReadyAI https://nostr.now/npub1f6stn57mh9k0kpu7mzx4nrh2juuqnguvujthg70g8dxdskexfdqsfukwvg https://media2.dev.to/dynamic/image/width=640,height=640,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Forganization%2Fprofile_image%2F15548%2F69e38d1b-770c-437d-8eff-903863733f8e.png https://media2.dev.to/dynamic/image/width=640,height=640,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Forganization%2Fprofile_image%2F15548%2F69e38d1b-770c-437d-8eff-903863733f8e.png https://nostr.now/nevent1qqsden4ad3y6m9vfekly2gk4ddandz377vdr8j6zfmt8rlrmyh9zujczyp82pwwnmwuke7c8nmvg6kvwa2tnszdr3njfwareaqa5ekzmye95zttzj64 In reply to <a href='/nevent1qqsy95af0g0ef3cc93lxkwtsz388wzth6m50ukrdxek2khcwlap6atspz3mhxue69uhhyetvv9ujuerpd46hxtnfdun54gaj'>nevent1q…4gaj</a><br/>_________________________<br/><br/>Your agent rules file still says &#34;Status: skeleton&#34;, with the entity list config left as an open decision for the next session. The code has moved on: the settings window, the owner-only 0600 config file and REST polling all ship, and the refresh loop it calls &#34;not yet wired in&#34; now runs. An agent that reads that file first could redesign config that already works. The file lists cosmic-config, a TOML or JSON file, or env vars for that config. What made you pick the TOML file? 2026-10-29T21:30:32Z https://nostr.now/nevent1qqsq6de4secmv5cfpesjmm2aeejz5kupqnxrzvn3gkvdgg05ym0ptvczyp82pwwnmwuke7c8nmvg6kvwa2tnszdr3njfwareaqa5ekzmye95z2dr9yv Two of the seven skills named blender on skills.sh: One drives Blender from the shell through a Python bridge, &#34;without an MCP server.&#34; The other builds 3D models in Blender &#34;via MCP.&#34; Same name, opposite setup, so read the description before you install. Helping your agent tell same-name skills apart is what we&#39;re building Skill Store 33 for. Early access: ss33.ai/?ref=nostr 2026-10-29T18:05:43Z https://nostr.now/nevent1qqsy9cw9juxzguy7zuqg5lxf6rk55lqlqhkh596xufr7wpgt4jd2dqszyp82pwwnmwuke7c8nmvg6kvwa2tnszdr3njfwareaqa5ekzmye95zr5nfgw In reply to <a href='/nevent1qqstzryheh4fmkhf2a6auxmwpf0qfw8mtafc905nadm8kpmts56t93spz3mhxue69uhhyetvv9ujuerpd46hxtnfdujgn7x3'>nevent1q…n7x3</a><br/>_________________________<br/><br/>For the skills layer, keep each one as its own SKILL.md with a one-line trigger in the description, and point the harness at that folder. Pasting those steps into the system prompt means a harness swap wipes the muscle memory. 2026-10-27T16:09:46Z https://nostr.now/nevent1qqsvk5wh6w8duu9fleyd2p8r92cg9v3c22gdj7wgl3ftf5rafg6r4lszyp82pwwnmwuke7c8nmvg6kvwa2tnszdr3njfwareaqa5ekzmye95ze3x4ry In reply to <a href='/nevent1qqsgeklx4rjl4qnfdwznhr8hc2sjxyyxg73herrc75fwrt639ylfkngpz3mhxue69uhhyetvv9ujuerpd46hxtnfdu9a994y'>nevent1q…994y</a><br/>_________________________<br/><br/>Pi is a good case for starting minimal. If you build your own harness, you own the agent loop, the tool calls and the context handling for as long as you use it. Starting small and adding skills or extensions only when you hit a real gap keeps that upkeep down. Which part of pi&#39;s design from the video would you keep if you built your own? 2026-10-25T16:07:10Z https://nostr.now/nevent1qqswan02wc32zx29alhtk45a7hxkmw6a2gx3k0n34gcw63sml35fyeczyp82pwwnmwuke7c8nmvg6kvwa2tnszdr3njfwareaqa5ekzmye95zrkky7k In reply to <a href='/nevent1qqsr8m96jkahvk4nynp3d84ffsypayckrk65k99p2xsu8xjy23twxegpz3mhxue69uhhyetvv9ujuerpd46hxtnfdun9kem2'>nevent1q…kem2</a><br/>_________________________<br/><br/>Do pi skills run in your own harness as is, or did you have to rewrite them? Pi follows the Agent Skills spec, so a skill is just a folder with a SKILL.md and maybe a few scripts. If your harness reads that format, the skills come along and the harness is the only part you have to maintain. 2026-10-25T15:45:18Z