Brag Sheet: Crafting Evidence-Backed Impact Statements for Performance Reviews from Git, GitHub, and Copilot Session Logs

1. Ask the agent to document your accomplishments, for example: "What did I accomplish in the last two weeks? I want to prepare my self-assessment." You can also mention a specific category such as infrastructure. 2. The agent confirms the time range and scope with you before scanning. It then checks the available tools: git, gh, and extension tools such as save_to_brag_sheet, and whether the ~/.copilot/session-state/ directory exists. 3. The agent scans the available sources: your commits in the current repository, merged pull requests via gh pr list, and workspace.yaml files in Copilot sessions. After that, it groups related work into a single entry. 4. The agent drafts entries in the "Action → Result → Evidence" format and classifies them into categories such as pr, bugfix, and oncall. Any entry lacking evidence is marked with the phrase (evidence needed) and is not deleted. 5. Review and edit the drafts. You can request a "review packet" that selects the top 3 to 5 accomplishments, organizes them by impact themes, and writes STAR-style narratives for them. 6. You will receive Markdown-formatted text grouped by week or by theme, ready to paste into your review document.

- It can work with no dependencies at all via the guided interview. - Optional tools: - git to extract commits from the current repository. - The gh tool with a signed-in GitHub account to fetch pull requests. - GitHub Copilot CLI session logs in ~/.copilot/session-state/, which are only available to those who have used Copilot CLI. - The save_to_brag_sheet, review_brag_sheet, and generate_work_log tools, which are only available after installing the separate copilot-brag-sheet extension. The agent must not call them unless it has confirmed they are available. - Limitations: - The skill does not invent metrics or numbers. - It does not capture work that leaves no trace in git, such as mentoring and design documents, except through the interview. - Commit scanning is limited to the current repository unless otherwise requested. - The skill is designed primarily for the Copilot CLI environment, and the scanning part requires an agent capable of executing terminal commands.