jgs-upstream-feedbacklisted

<!-- Copyright (c) 2026 JG Systems Consulting Ltd. Source: https://github.com/jgsystemsconsulting/jgs-archi-skills. See LICENSE. --> <!-- SPDX-License-Identifier: MIT --> # jgs-upstream-feedback ## When to use User-invoked, or dispatched by a pack orchestrator, when a shipped skill or MCP contract blocked an outcome that any user of the pack would hit. Not a modelling specialist. Does not talk to Archi. ## Prerequisites GitHub CLI (`gh`) is optional. Logged-in `gh` files the issue; otherwise this skill prints a pre-filled GitHub new-issue URL. Python is not required. ## Hard rules 1. File nothing until the user says yes to the exact draft, unless already_approved is true and the body is unchanged. 2. Skill-meta only. No element names, model paths, viewpoint content, tokens, or credentials. 3. Origin only if any user would hit this. Local model, org, install, or governance pain stays local. 4. No bot token. File as the logged-in user. 5. Never open a pull request against the MCP plugin. Never patch plugin source. 6. allow_pr stays false unless the user asked for a patch and the current workspace is a clone of the skill pack origin. 7. Do not phone home. Do not write a lessons file. Do not stall a modelling session on login. ## Caller payload Orchestrator or this skill (direct invoke) must have: - kind: skill-gap or mcp-gap or bug - issues_repo: owner/name - title: short, no model nouns - body: the markdown template below, already filled - labels: list, or empty - all