Repository navigation
Conversation
Repeating the coupled plugin checklist and lifecycle commands at the entrypoint requires maintaining a second policy copy. The entrypoint already requires reading both fuller authorities before any change, so removing these duplicates preserves the task actions without another checklist to keep aligned. The same audience, trigger, and duties remain in CONTRIBUTING.md lines 31-53 and AGENTS.md lines 10-24 at analyzed commit: https://github.com/github/plugin-marketplace-starter/blob/011cb1f8271b7af26b8301595422dbd48e2fe45a/CONTRIBUTING.md#L31-L53 https://github.com/github/plugin-marketplace-starter/blob/011cb1f8271b7af26b8301595422dbd48e2fe45a/AGENTS.md#L10-L24 The lifecycle rule retains scaffolding, regular and shipped-example removal, renaming, and registration handling. Self-contained packages and coordinated manifest, catalog, changelog, skill-path configuration, and eval changes remain required. The mandatory reads and repository-only audience are unchanged. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…list A repeated build-and-verify rule creates another place to maintain a procedure already loaded before changes. Removing the duplicate leaves the generated-source condition, generated-file deliverable, and build-before-verify order with the required maintenance rules and contributor checklist. The surviving conditional build rule is at AGENTS.md lines 41-45 and the ordered validation procedure is at CONTRIBUTING.md lines 63-74: https://github.com/github/plugin-marketplace-starter/blob/011cb1f8271b7af26b8301595422dbd48e2fe45a/AGENTS.md#L41-L45 https://github.com/github/plugin-marketplace-starter/blob/011cb1f8271b7af26b8301595422dbd48e2fe45a/CONTRIBUTING.md#L63-L74 These are already mandatory reads from the unchanged entrypoint. This cleanup does not change validation commands or permit omitting generated outputs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Maintaining the paid-eval authorization warning twice is unnecessary when the entrypoint already requires its more complete maintenance rule. Removing the duplicate leaves the credit warning, both paid commands, eval workflow, explicit user-request condition, and enablement restriction in that required authority. The surviving duty is at AGENTS.md lines 72-74: https://github.com/github/plugin-marketplace-starter/blob/011cb1f8271b7af26b8301595422dbd48e2fe45a/AGENTS.md#L72-L74 Keep the existing model-free command guidance as a standalone bullet: it still helps readers select safe inspection commands without running paid evals. No authorization or cost distinction is weakened. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
| read [AGENTS.md](../AGENTS.md) for the maintenance rules and | ||
| [CONTRIBUTING.md](../CONTRIBUTING.md) for the change checklist. | ||
|
|
||
| - Keep each plugin self-contained, and change its manifest, catalog entry, |
There was a problem hiding this comment.
🧹 One maintenance checklist: Repeating these duties requires keeping a second policy copy aligned even though this entrypoint already requires both authorities. The coupled plugin checklist retains self-contained packages and coordinated manifest, catalog, changelog, skill configuration, and eval changes; the complete lifecycle rule retains scaffolding, both removal routes, renaming, and registration handling. Commit
| changelog, `.vally.yaml` entries, and evals together. Use | ||
| `npm run new-plugin`, `npm run remove-plugin` (or `remove-example`), and | ||
| `npm run rename-marketplace` instead of editing registrations by hand. | ||
| - Run `npm run build` after changing generated-file sources, then `npm run verify`. |
There was a problem hiding this comment.
🧹 One validation procedure: A repeated build rule creates another place to maintain a procedure already required before changes. The generated-source condition and output duty and build-before-verify checklist remain in the mandatory reads, so this deletion does not relax validation. Commit
| - Don't run `npm run eval`, `npm run eval:smoke`, or the evals workflow unless | ||
| asked; they spend AI credits. | ||
| `npm run eval:list` and `npm run check-remotes` don't call a model. | ||
| - `npm run eval:list` and `npm run check-remotes` don't call a model. |
There was a problem hiding this comment.
🛠️ One paid-eval permission rule: Maintaining the authorization warning twice is unnecessary when the entrypoint already requires the fuller rule. The surviving credit warning and authorization condition retain both paid commands, the workflow, the user-request prerequisite, and the enablement restriction; the model-free inspection guidance remains available as its own bullet. Commit
There was a problem hiding this comment.
🟢 Approval recommended
The reviewed change has no active issues and preserves the required guidance references and model-free command notes while removing duplicated instructions.
No open issues 🎉
What changed in this PR
This pull request centralizes maintenance and contribution guidance in AGENTS.md and CONTRIBUTING.md, simplifying the repository’s Copilot entrypoint while preserving required references and model-free command notes.
Changes:
- Removed duplicated plugin maintenance, generated-file, verification, and paid-evaluation guidance.
- Retained required reads for
AGENTS.mdandCONTRIBUTING.md. - Preserved notes that
npm run eval:listandnpm run check-remotesdo not call a model.
| File | Description |
|---|---|
.github/copilot-instructions.md |
Simplifies the entrypoint while preserving required guidance references and model-free command information. |
Review effort: Lite (auto)
|
@copilot resolve the merge conflicts in this pull request |
…memories-20261106153300 # Conflicts: # .github/copilot-instructions.md Co-authored-by: ronniegeraghty <28957151+ronniegeraghty@users.noreply.github.com>
Resolved the conflict by accepting |
🧠 How this helps Copilot
Copilot already has to read the maintenance rules and contributor checklist before changing this repository. Keeping plugin maintenance, generated-file validation, and paid-eval permission duties in those required owners avoids maintaining parallel policy copies at the entrypoint.
The mandatory reads, model-free inspection commands, and repository-only scope remain, so the shorter entrypoint preserves the guidance needed to follow those workflows.
Automated details
Source:
curate-memoriesDream, analyzingmainat011cb1f.