Context
Currently, I'm wrapping beans in complex bash/jq scripts to derive the "next" task or "audit" data integrity against git history for a project I'm working on. As I had the agent work on optimizing it (it's extremely slow), I asked whether we should recommend any upstream features, and here's what it strongly suggested.
Proposal:
beans next: Implement native task ranking in Go. Use the existing tree-based hierarchy to traverse the task list. It can identify "ready" leaf nodes, prioritize items based on their position in the tree or metadata (like priority), and filter out blocked/completed work instantly.beans audit: Add a command to validate bean integrity. This would check for broken tree structures (e.g., orphaned children) and optionally use go-git to scan the repository's commit history for bean IDs to flag "stale" open beans that have already landed on the main branch.
Benefit
Native Go implementation provides machine-optimized, zero-dependency endpoints for AI agents to decide their next move and verify their work, staying consistent with the project's "agent-first" philosophy.
Context
Currently, I'm wrapping beans in complex bash/jq scripts to derive the "next" task or "audit" data integrity against git history for a project I'm working on. As I had the agent work on optimizing it (it's extremely slow), I asked whether we should recommend any upstream features, and here's what it strongly suggested.
Proposal:
beans next: Implement native task ranking in Go. Use the existing tree-based hierarchy to traverse the task list. It can identify "ready" leaf nodes, prioritize items based on their position in the tree or metadata (like priority), and filter out blocked/completed work instantly.beans audit: Add a command to validate bean integrity. This would check for broken tree structures (e.g., orphaned children) and optionally use go-git to scan the repository's commit history for bean IDs to flag "stale" open beans that have already landed on the main branch.Benefit
Native Go implementation provides machine-optimized, zero-dependency endpoints for AI agents to decide their next move and verify their work, staying consistent with the project's "agent-first" philosophy.