Acontext vs skill-recorder
Skill memory layer for agents: auto-captures learnings from runs into plain Markdown skill files you can read, edit, git and share across frameworks — memory without an opaque store. — versus — Desktop app that records a real work session — clicks, apps, pages, narration — and uses Copilot CLI to reconstruct intent + steps, then generates a reusable SKILL.md or scheduled Automation.
Same artifact from different evidence: Acontext distills skills from agent session logs; Skill Recorder distills them from watching the HUMAN work. Both land as SKILL.md.
| Acontext | skill-recorder | |
|---|---|---|
| Stars | 3.7k | 2.5k |
| Forks | 332 | 267 |
| Language | JavaScript | TypeScript |
| License | Apache-2.0 | MIT |
| Last activity | 26 days ago | 2 days ago |
| Topics | memory, skills | skills |
| Curated connections | 6 | 2 |
Acontext — the curator's take
The "memory should be legible" bet: instead of embeddings in a vector store, learnings from agent runs become Markdown skill files you can read, diff, git and mount into any framework — debuggable memory users can inspect and correct, the exact thing opaque memory layers get wrong. It can also adopt and evolve skills you wrote or downloaded. Trade-off: no semantic recall over thousands of entries; it lives or dies on distilling runs into a curated, manageable skill set. Want scale-out retrieval memory instead? That's memmachine or hindsight territory.
skill-recorder — the curator's take
Demonstration beats description: recording yourself doing a task once captures the tacit steps you'd never think to write down, and generalizing to native tools (gh, web_fetch) instead of replaying clicks is the right call. When NOT: it requires GitHub Copilot access and its Automation output targets Microsoft's agent surfaces (Scout, Cowork, Copilot Studio) — the SKILL.md is portable, the rest of the pipeline isn't; screen recording means trusting local capture with everything on screen.