Call for public or synthetic agent workflows to audit with LoopX

[huangruiteng]

I am looking for public or synthetic long-running AI agent workflows that LoopX
can audit.

LoopX is a control layer for long-running AI agents. For this first pass, I want
one workflow where the agent technically runs, but cost, retries, review
latency, unclear gates, stale state, or owner decisions are messy.

What I can return:

• a one-page decision card;
• where time or money was wasted;
• which evidence is trustworthy;
• which decision is needed next;
• what to continue, narrow, or stop.

What I do not need:

• private data;
• credentials;
• raw private logs;
• production access;
• customer data.

Useful examples:

• an issue-to-PR coding-agent run that required too much review;
• a multi-agent task that retried tools or repeated work;
• an agent PoC that worked in demo but was hard to judge;
• an observability trace that shows what happened but not what to do next;
• a synthetic workflow that represents a real cost/review problem.

If you have one, please comment with:

1. the public repo, issue, trace, or synthetic task;
2. what felt wasteful or hard to review;
3. what decision you wish a workflow owner could make faster.

Success for this issue is one concrete workflow that can become a metadata-only
LoopX audit. If no useful workflow appears, I will close the issue and try a
more specific channel.

[huangruiteng]

Discussion mirror for broader workflow sharing: #673