◈—Why this repository exists
Ideas, not implementations
The Cognitive Backlog is the long-term backlog of cognitive capabilities for the Agent Operated Environment. Each issue is a desired capability. Capabilities are implementation-agnostic — a capability may eventually become any of these:
loopagentMCP toolworkflowdashboardintegrationanother reusable component
LoopForge consumes issues from this repository and transforms them into executable implementations.
— README.md
Priority answers urgency. Cognitive Value answers compounding intelligence.
- P0 Blocking, urgent, or required for the next operational step.
- P1 High-value and near-term. Should be actively considered for forging.
- P2 Worth keeping warm. Needs analysis, shaping, or sequencing.
- P3 Interesting but not currently important. Archive later if it stops earning its keep.
- ⭐ Local convenience. Useful, but narrow.
- ⭐⭐ Improves one workflow or recurring task.
- ⭐⭐⭐ Improves a system area or class of tasks.
- ⭐⭐⭐⭐ Compounds across multiple workflows or agents.
- ⭐⭐⭐⭐⭐ Strategically changes what the Agent Operated Environment (AOE) ecosystem can do.
High cognitive value does not always mean high priority. A capability can be transformative but blocked by missing infrastructure, unclear evaluation, or too much blast radius for the current phase.
What this repository deliberately does not do
compile loopsgenerate loop promptsrun LoopForgeexecute agentsschedule tasksmonitor systemsanalyze runtime resultsreplace Codex
Require audience-bound OAuth tokens for /mcp; accepting tokens minted for some other API is confused-deputy nonsense with better branding.
— docs/chatgpt-publisher.md