Turns vague "what did I do?" into evidence-backed impact statements for performance reviews, self-reviews, promotion packets, and weekly updates. Uniquely mines Copilot CLI session logs to reconstruct forgotten work, plus git commits and GitHub PRs. Enforces a 3-part impact contract (action → result → evidence). Works standalone with one runtime dependency: the official MCP SDK (when running the MCP server). Use this skill whenever the user asks to document, summarize, reconstruct, or organize their engineering work — even if they don't literally say "brag sheet". Trigger phrases include: "brag", "log work", "what did I do", "backfill my work history", "performance review", "self-review", "self assessment", "write impact statement", "review prep", "promo packet", "promotion case", "weekly update", "status report", "accomplishments", "what did I ship", "I forgot to log my work", "summarize my work", "track my wins", "end of half", "career growth", "work journal".
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Turns vague "what did I do?" into evidence-backed impact statements for performance reviews, self-reviews, promotion packets, and weekly updates. Uniquely mines Copilot CLI session logs to reconstruct forgotten work, plus git commits and GitHub PRs. Enforces a 3-part impact contract (action → result → evidence). Works standalone with one runtime dependency: the official MCP SDK (when running the MCP server). Use this skill whenever the user asks to document, summarize, reconstruct, or organize their engineering work — even if they don't literally say "brag sheet". Trigger phrases include: "brag", "log work", "what did I do", "backfill my work history", "performance review", "self-review", "self assessment", "write impact statement", "review prep", "promo packet", "promotion case", "weekly update", "status report", "accomplishments", "what did I ship", "I forgot to log my work", "summarize my work", "track my wins", "end of half", "career growth", "work journal".
license
MIT
compatibility
Cross-platform (Windows, macOS, Linux). Works with any GitHub Copilot CLI session. Optional: git, gh CLI.
metadata
{"version":"1.1"}
argument-hint
Optional: time range ("last 2 weeks", "this half"), category ("infrastructure"), "backfill", or "review prep"
Brag Sheet — Work Impact Writer
Turn engineering work into evidence-backed impact statements for performance reviews, self-reviews, promotion packets, and weekly updates. Uniquely mines Copilot CLI session logs, git history, and PRs to reconstruct forgotten work.
USE FOR: "brag", "log work", "what did I do", "backfill", "performance review", "self-review", "promo packet", "weekly update", "status report", "write impact statement", "what did I ship", "I forgot to log my work", "review prep", "accomplishments"
DO NOT USE FOR: project management, sprint planning, time tracking, ticket creation
Quick Start
User wants...
Mode
Output
Log one accomplishment
Capture
1 impact-first entry
"What did I do last week?"
Backfill
Entries grouped by week, mined from git/PRs/sessions
Prep for review or promo
Review Pack
Entries grouped by impact theme + STAR narratives
Agent Behavior Rules
DO confirm the time range and scope before scanning sources. Don't assume "last week" — ask.
DO check which tools are available (save_to_brag_sheet, git, gh) before choosing a workflow.
DO always include all three parts: action → result → evidence. If evidence is missing, write (evidence needed) — never silently omit.
DO show drafted entries to the user before saving. Never auto-save without confirmation.
DO group related commits into a single entry. Ten commits on the same feature = one entry.
DO preserve the user's voice. Reframe for impact, but don't invent accomplishments or inflate scope.
DO NOT fabricate metrics, team sizes, or impact numbers. If the user doesn't provide a number, don't invent one.
DO NOT write entries for work the user only described verbally without verifying. Ask: "Did this ship? Is there a PR or doc I can reference?"
DO NOT skip the backfill scan steps or draft entries before scanning is complete.
DO NOT pad weak periods with trivial entries. An honest gap is better than inflated fluff.
Entry Format
Every entry uses impact-first framing with three required parts:
Did [action] → [result/impact] → [evidence]
Do not output an entry unless it includes all three parts. If evidence is missing, ask for it or mark as "(evidence needed)".
Anti-Patterns
❌ Don't
✅ Do instead
"Fixed a bug in auth"
"Fixed token refresh race condition → eliminated 401s affecting 12% of API calls → PR #247"
"Worked on dashboards"
"Built latency dashboard in Grafana → on-call detects P95 spikes in <2min → deployed to prod"
Invent a metric: "saved 40% of eng time"
Ask: "Do you have a rough estimate, or should I keep this qualitative?"
One entry per commit
Group related commits into one entry with highest-impact framing
Passive voice: "The pipeline was improved"
Active voice: "Built CI matrix → caught Windows-only bug before release"
List technologies used
State the outcome: "Migrated 4 services to IaC → deploy time 45min → 8min"
Silently drop weak entries
Mark (evidence needed) and present for user to fill in
Evidence Ladder
Not every entry needs a metric. Use the strongest evidence available:
Strength
Evidence type
Example
🥇 Best
Quantified metric
"Reduced P95 latency from 800ms to 120ms"
🥈 Strong
PR, commit, or doc link
"PR #312, design doc in wiki"
🥉 Good
Observable outcome
"Unblocked Team X", "Resolved Sev2 incident Y"
✅ Acceptable
Qualitative + context
"Reduced toil for on-call rotation — see updated runbook"
⚠️ Weak
Activity only
"Worked on auth" — reframe or mark (evidence needed)
Never invent a metric to fill the gap. Qualitative evidence with context beats fabricated numbers.
Categories
ID
Emoji
Use for
pr
🚀
Merged PRs, shipped features
bugfix
🐛
Bug fixes, incident patches
infrastructure
🏗️
Infra, deployments, migrations
investigation
🔍
Root cause analysis, debugging
collaboration
🤝
Reviews, mentoring, design discussions
tooling
🔧
Dev tools, scripts, automation
oncall
🚨
Incident response, on-call wins
design
📐
Design docs, architecture decisions
documentation
📝
Docs, runbooks, guides
How to Help the User
Follow this decision tree:
If save_to_brag_sheet tool is available → use extension tools directly (save_to_brag_sheet, review_brag_sheet, generate_work_log). Do not reference or attempt to call these tools unless they are confirmed available.
If git or gh CLI is available → backfill from commits and PRs (see Backfill section below)
Otherwise → guided interview: "What did you work on?", "Who benefited?", "What's the evidence?"
For each entry, walk through: What (the deliverable) → Why (who benefits) → Evidence (PR, metric, link). Output formatted markdown the user can paste into a review doc.
Backfill Workflow
When the user asks "what did I do last week" or "backfill my history":
Follow these steps in order. Do not draft entries until scanning is complete.
Step 1: Scan available sources
Check what's available, then mine each source:
git --version 2>/dev/null # for commit mining
gh --version 2>/dev/null # for PR miningls ~/.copilot/session-state/ 2>/dev/null # Copilot session logs
Git commits — recent commits by the user in the current repo:
Ask for gaps — if evidence is missing, prompt the user: "What metric changed?", "Who was unblocked?", "What's the PR or incident ID?"
Strong vs weak entries
✅ Strong
❌ Weak
Outcome-first, quantified
Activity list ("worked on X")
Tied to customer/team impact
No beneficiary mentioned
Includes evidence (PR, metric)
No measurable result
Shows ownership or leadership
Pure task completion
Narrative format
For longer narrative sections, use STAR: Situation → Task → Action → Result.
For Microsoft employees using the Connect preset, frame entries around Core Priorities: delivering results, customer obsession, teamwork, and growth mindset.
Output Contract
Before finishing, ensure:
Every entry has action → result → evidence (mark (evidence needed) if missing)
No fabricated metrics — only user-provided or source-verified data
Entries shown to user before saving
Time range explicitly stated
Output is pasteable markdown with categories assigned
Gotchas
No recent commits in the current repo
The user may work across multiple repos. Before concluding there's nothing to backfill:
Ask if they want to scan a different repo or branch
Check gh pr list --author @me --state merged for cross-repo PRs
Fall back to the guided interview — not all impactful work leaves git traces (design docs, incident response, mentoring)
Review period doesn't match git history
Performance reviews often cover 6–12 months. Explicitly set the date range:
PR history (gh pr list --state merged) is more reliable for long time ranges than commit logs.
User can't quantify impact
Not every entry needs a number. See the Evidence Ladder above. Acceptable evidence includes PR links, "unblocked Team X", or qualitative outcomes with context. Never invent a metric to fill the gap.
Copilot session directory doesn't exist
~/.copilot/session-state/ only exists if the user has run Copilot CLI sessions. Don't error — silently skip and note: "No Copilot session history found; scanning git and PRs only."
"brag" might mean something else
The user might say "brag about this feature to my team" (a launch announcement, not a work entry). Confirm intent if ambiguous.
Pair programming or co-authored commits
If multiple authors appear on the same commits, ask: "Should I credit this as your work, shared work, or skip it?"
Automatic Session Tracking (Optional)
For automatic background tracking of every Copilot CLI session (files edited, PRs created, git actions), install the copilot-brag-sheet extension. It adds save_to_brag_sheet, review_brag_sheet, and generate_work_log tools to every session.
