dpd-feedbacklisted

# /dpd-feedback — Submit dogfood feedback as a GitHub issue Announce: "Using dpd-feedback skill." ## Purpose Surface friction / bugs / feature suggestions you noticed while using DPD, as GitHub issues on the canonical repo. Cheap, low-friction path: `/dpd-feedback "<short description>"` produces a pre-filled issue draft, you confirm, it gets submitted. ## Privacy hard constraint (NON-NEGOTIABLE) **This skill MUST NOT send any session content under any circumstances.** This includes: - Node text (any of it — no decision text, hypothesis text, evidence text, rationale text, ANY node text) - Scope name - Session label - Session id (correlation key, exclude) - Pool item content - Rejected hypothesis text - Mermaid / YAML / markdown exports of the graph Allowed payload: 1. **User free-text description** (the `<short description>` argument and any follow-up the user types when prompted to expand) 2. **System metadata only**: - DPD plugin version (read from plugin.json) - `dpd-mcp-server` Python package version (`pip show dpd-mcp-server`) - Agent name (Claude Code / Cursor / Codex / etc. — detect from env) - Python version (`python3 --version`) - OS name + version (`uname -a` minimized to OS family) - Optional: stack trace if the user opted to attach one for an error case If the user wants to attach graph context for a bug report, the skill MUST ask them to manually copy-paste the specific part they choose to share. **Tool-call policy (allow-list, NOT den