suggest-awesome-github-copilot-skills

This skill analyzes the current repository and recommends relevant Agent Skills from the GitHub awesome-copilot repository that are not already installed locally, while also flagging locally installed skills that have drifted out of date versus their upstream versions. It solves the discovery and maintenance problem of keeping a repository's `.github/skills/` folder aligned with a large, evolving upstream catalogue.

The process fetches the upstream skills list, scans local skill folders, reads front-matter names and descriptions, and fetches each corresponding upstream SKILL.md via raw GitHub URLs to compare content. It classifies local skills as up-to-date, outdated (with the specific differences in front matter, instructions, or bundled assets), and identifies candidate new skills by matching repository patterns (languages, frameworks, project types, CI/CD needs) and chat-history context against the catalogue. Results are presented as a structured comparison table with rationale and install status. Crucially, the skill waits for explicit user approval before installing or updating anything. On request, it downloads new skills and refreshes outdated ones into `.github/skills/`, preserving folder structure and pulling both SKILL.md and bundled assets, using the fetch tool and, where needed, curl to ensure complete retrieval, without altering file content.

It targets developers and teams who curate a local Copilot skills library and want to keep it comprehensive and current. Use cases include finding skills relevant to a project's stack, detecting outdated local copies, and updating them to the latest upstream version. Security-wise, it downloads files from the trusted github/awesome-copilot repository over raw GitHub URLs and requires explicit user consent before writing anything — it does not execute downloaded code or pipe remote content into a shell, so it rates as low risk despite using curl for retrieval.