Public intake point for feedback about the claude-agent-team
Claude Code plugin and the methodology it executes. Issues filed here are triaged
by the owner and promoted into the private source repositories.
⚠️ This repository is world-readable. Never file client-confidential content here — code, names, paths, credentials, or anything that identifies a client or engagement. Feedback should be about the plugin and methodology only.
- Automatically, via the plugin. The plugin's
/file-feedbackcommand scrubs an approved promotion candidate, previews it for a human, and — on explicit confirm — files it here using the operator's own GitHub auth. Such issues carry a machine marker and are labeledorigin:agent-team. - By hand, via the form. Anyone can open the Feedback issue form. Blank issues
are disabled so every submission is structured. Hand-filed issues are labeled
origin:unverified(they carry no plugin marker) — not a judgment, just a note that the owner should confirm the origin during triage.
Labels are applied automatically by the Auto-triage feedback Action on issue open;
non-collaborators cannot set labels themselves.
| Label | Meaning |
|---|---|
origin:agent-team |
Filed by /file-feedback with a valid provenance marker. |
origin:unverified |
No valid marker (hand-filed, or a marker that didn't validate). Review the origin. |
triage |
Awaiting owner triage. |
source:engagement-retro |
Field feedback bundled from an engagement retro. |
standard-feedback |
Routes to ai-assisted-dev-standard. |
prompt / process / positive-signal |
Methodology feedback class. |
This repo is Gate 1 of a two-gate writeback model. Gate 2 — promotion of accepted
feedback into the private source repositories — is owner-run. See
claude-agent-team ADR-0004
for the full design and the provenance-marker contract.