Where This Idea Came From

‍I asked Claude for a rundown of which Fulcra skills I could should start using today, and pointed it at the agent-skills page. It came back with a breakdown of the library sorted against how I actually use it and flagged two skills as the highest immediate payoff: fulcra-memory and fulcra-situational-awareness.

‍

‍Problem

‍Everything Claude learned about my ongoing threads only lived inside whatever chat we were having. Decisions about the renovation, the state of the PR work, the shape of my creative projects — none of it survived past that conversation's window, and none of it was visible to any other agent or session I might use later. There was also no way for Claude to notice, at the start of a new conversation, that anything had actually changed since the last one. Every session started from zero.

‍

‍Steps to Solve

1. Asked Claude which Fulcra skills were worth adopting given my actual usage; it triaged the full library into "most relevant now," "good but setup-dependent," and "not needed yet," rather than just listing everything on the page.
2. Had Claude read the  fulcra-memory skill spec and its MCP-tool reference before doing anything, since it's operating through the MCP connector rather than a CLI.
3. Directed Claude to create the OKF-compliant namespace at agent/claude/ in my Fulcra file store: a role.md describing what it is and how it operates, a progress.md for recent work and next steps, and the index.md and log.md that OKF requires for any directory.
4. Reviewed drafts of role.md and progress.md which showed me the exact content it planned to upload.
5. Once fulcra-memory was live, had Claude install fulcra-situational-awareness
6. Had it record that consent directly inside agent/claude/role.md, then run its first awareness scan with get_data_updates on the spot, so I could see the habit  working before trusting it to run unsupervised in future sessions.

‍

Solution

Claude now keeps its own working memory in my Fulcra account instead of in chat history . Every setup decision and next step gets written to agent/claude/progress.md and log.md, and role.md documents both what Claude is for and the standing habits it's agreed to, including scanning for changes at the start of a session.The whole thing conforms to OKF, so any other agent I connect to this Fulcra account later can read agent/claude/ without me re-explaining anything.

‍

‍Result

‍The next new conversation I start with Claude checks Fulcra for what's changed — new files, freshly ingested data — before diving into a task, and it can read its own progress.md to pick up exactly where the last session left off. The bigger payoff shows up the moment I bring in a second agent on a different platform because the memory lives in Fulcra rather than inside one chat history, that agent inherits the same context automatically instead of me re-explaining the PR work, the creative projects, and the renovation status all over again.