Help Needed

Open questions and open roles — where outside perspective or expertise would change things.

People

Skills and backgrounds we'd love to have in the collective. Not roles — just people who'd fit.

Decisions

Open questions where we need input before making a call.

Sys Prompt User Msg Agent Tool Call Session archive — every turn persisted not scored today Offline Evals fixtures → expected tool LLM-as-Judge score real transcripts Today: dev testing + UAT catch what they catch
help needed · 22 Apr 2026

How Do We Know Agents Pick the Right Tool?

Agent tool choice is verified by dev testing and UAT. The raw data to do better is already being collected — we just aren't scoring it.

opentestingagents
Skills & Workflows (self-improving) sharpen triggers · prune unused steps improves activates Memory (cross-project · self-curating) promote · freshen · rank · prune signal in context out Sessions (Claude · Copilot · agents at work) usage · outcomes · incidents · reranker scores
help needed · 21 Apr 2026

Memory as Substrate

What I've learned building a memory system for agents, the vision for the self-improving skill layer it's meant to enable, and where I'd welcome deep help.

openmemoryarchitecture
upstream/main agent-purple-shovel fork/main agent-beehave DEVIATIONS.md 45+ files tracked · grows every sprint T1 Full isolation Ignore upstream Low risk T2 Additive ext. Re-add import Med. risk T3 Inline patches Reapply on sync High risk
help needed · 22 Mar 2026

Fork Governance at Scale

DEVIATIONS.md is tracking 45+ differing files and growing — at what point does a fork become a liability?

openupstreamgovernance
Agent A Agent B Agent C Community Pool shared · sharedWithId:"*" Quality Gate who approves shared memories? ? Memory Decay should old memories expire? ? Contradictions agents remember opposite things ?
help needed · 22 Mar 2026

Shared Memory Quality

Agent Amber's community pool learns from everyone's interactions — but what stops bad memories from poisoning it?

openmemorycommunity
Vanilla Monitor local machine → SSH → Production 143.110.238.159 container-health api-health upstream-drift usage-anomalies Slack DM ⚠ Laptop may sleep monitor goes dark silently
help needed · 22 Mar 2026

The Vanilla Production Monitor

We use an upstream instance to monitor production — it works, but is it the right long-term answer?

openmonitoringinfrastructure
Agents Prototyping Memory CI/CD Humans Prod Deploy Rollback Monitoring push the line →
help needed · 21 Mar 2026

The Automation Boundary

What agents handle today, what still needs humans, and where we want to push the line.

openautomation
Deploy Agent Build Agent Test Agent Event Bus (pub/sub) Daniel Nathan Sarah
help needed · 21 Mar 2026

Cross-Agent Communication

How should agents working on different people's projects talk to each other?

open
GH Runners ? Convex Hosting ? Monitoring ? Scaling ? Docker Proxy ? Auto-Deploy ?
help needed · 21 Mar 2026

Infrastructure Decisions

Six open infrastructure questions — GH runners, Convex hosting, monitoring, scaling, and more.

openinfrastructure
Opt 1: Project ACL recommended Opt 2: Workspaces Opt 3: Fine ACLs
help needed · 21 Mar 2026

Privacy Controls

How should personal vs shared projects work? Separate logins, visibility settings, or something else?

open