blameprompt

March 20, 2026 · View on GitHub

Your AI skills deserve a portfolio.

Track your AI coding activity across 15 agents. Build your developer score. Earn achievement badges. Climb the leaderboard. Get a public profile at blameprompt.com/username that proves your AI coding skills — with real data, not self-reported claims.

$ blameprompt profile

  Username:  alice
  Email:     alice@dev.com
  Score:     82 (Expert)
  Rank:      #12 on leaderboard
  Badges:    8/12 unlocked
  Prompts:   14,832
  Streak:    47 days
  Profile:   blameprompt.com/alice

Install

# macOS / Linux
curl -sSL https://blameprompt.com/install.sh | bash

# Windows (PowerShell)
irm https://blameprompt.com/install.ps1 | iex

# Or build from source
cargo install --path .

The installer automatically configures hooks for all 15 supported agents, installs Git hooks globally, and sets up the transparent git wrapper — every repo is tracked from that point forward.

Get started

# 1. Install (one-time)
curl -sSL https://blameprompt.com/install.sh | bash

# 2. Log in with GitHub
blameprompt login

# 3. Sync your data and build your profile
blameprompt sync

# 4. Open your dashboard
blameprompt dash

That's it. Your profile, score, heatmap, and badges start building automatically as you code with AI.

Or sign in directly at blameprompt.com/login.

What you get

Developer score (0-100)

A composite rating based on real data:

  • Prompt quality — clarity, actionability, context, and efficiency scoring
  • Acceptance rate — how much AI-generated code you keep
  • Cost efficiency — results without burning tokens
  • Model diversity — knowing when to use Opus vs Sonnet vs GPT-4o
  • Consistency — daily streaks and sustained usage

Ratings: Novice → Intermediate → Advanced → Expert → Elite

Public profile

Your profile at blameprompt.com/username showcases:

  • Developer score and rating
  • GitHub-style prompt heatmap
  • Achievement badges (12 types)
  • Model proficiency breakdown
  • Language, tool, and editor usage
  • AI vs human code ratio

Put it on your resume. Embed a badge in your GitHub README. Share on LinkedIn.

Achievement badges

12 badges that track real milestones:

  • Century — 100 prompts
  • Thousand Club — 1,000 prompts
  • Week Warrior — 7-day streak
  • Month Master — 30-day streak
  • Polyglot — 5+ programming languages
  • Multi-Agent — 3+ AI agents used
  • Cost Pro — below-average cost per prompt
  • Elite — developer score 90+
  • And 4 more to discover

Global leaderboard

Rank against developers worldwide by prompts, score, or active days. Browse the leaderboard at blameprompt.com/leaderboard.

How it works

AI Agent ──> hooks/import ──> staging.json ──> git commit ──> Git Notes ──> sync ──> profile
              (files, lines,     (local,        (post-commit    (refs/notes/    (score, badges,
               prompt, model,     gitignored)     hook)          blameprompt)    heatmap, rank)
               tools, cost)
  1. You code with AI — hooks fire automatically or you import sessions from other agents
  2. One receipt per prompt — files, lines, model, cost, tokens, quality score — all captured
  3. Receipts attach on commitpost-commit hook writes everything as a Git Note
  4. Sync to your profileblameprompt sync uploads aggregated metrics to your public profile
  5. Score and badges update — your developer score, heatmap, and achievement badges build automatically

Everything is local-first. Nothing leaves your machine unless you choose to sync.

Supported agents (15)

All detected agents are auto-configured by blameprompt init --global. If an agent isn't installed, it's silently skipped.

AgentHook configImport historical sessions
Claude Code~/.claude/settings.jsonAutomatic
GitHub Copilot~/.github/hooks/blameprompt.jsonblameprompt record-copilot
OpenAI Codex CLI~/.codex/config.tomlblameprompt record-codex
Google Gemini CLI~/.gemini/settings.jsonblameprompt record-gemini
Cursor~/.cursor/hooks.jsonblameprompt record-cursor
Windsurf (Codeium)~/.windsurf/hooks.jsonblameprompt record-windsurf
Antigravity IDE~/.antigravity/settings.jsonblameprompt record-antigravity
Continue~/.continue/hooks.jsonblameprompt record-continue
Droid~/.droid/hooks.jsonblameprompt record-droid
JetBrains Junie~/.junie/hooks.jsonblameprompt record-junie
Atlassian Rovo Dev~/.rovo-dev/hooks.jsonblameprompt record-rovo-dev
Sourcegraph Amp~/.amp/hooks.jsonblameprompt record-amp
OpenCode~/.opencode/hooks.jsonblameprompt record-opencode
Any providerblameprompt record --session <file> --provider <name>

VS Code extension

Install from VS Code Marketplace.

  • Prompt Receipts — tree view of commits > prompts > files/tools/agents
  • Prompt History — git-log-style timeline with model-colored labels
  • File History — all prompts that modified the currently open file

Commands

Account & profile

blameprompt login                   # sign in via GitHub (opens browser)
blameprompt login --token <key>     # authenticate with API token (CI/headless)
blameprompt logout                  # clear stored credentials
blameprompt profile                 # show your score, rank, and badges
blameprompt profile --edit          # edit profile in browser
blameprompt sync                    # sync receipts to your public profile
blameprompt dash                    # open dashboard in browser

Attribution

blameprompt blame src/auth.rs       # line-by-line AI vs human
blameprompt show a1b2c3d            # all receipts for a commit
blameprompt search "JWT"            # full-text search across prompts
blameprompt diff                    # annotated working-tree diff
blameprompt diff a1b2c3d            # annotated commit diff
blameprompt check-provenance src/auth.rs          # AI vs human lines
blameprompt check-provenance src/auth.rs --line 5 # specific line

Analytics & reporting

blameprompt analytics                       # AI score, cost breakdown, model usage
blameprompt audit                           # full audit trail (md, table, json, csv)
blameprompt audit --from 2026-01-01 --author "Jane" --format json
blameprompt report --output report.md       # comprehensive markdown report
blameprompt report --quality                # prompt quality scoring report

Security

blameprompt vuln-scan               # CWE pattern scanning on AI-generated code
blameprompt prompt-injection        # detect backdoors and hidden instructions
blameprompt secret-rotation         # flag secrets exposed to AI
blameprompt supply-chain-risk       # risk score 0-10
blameprompt license-scan            # model license compliance

Hackathon fairness

blameprompt hackathon-report                    # last 24h, all participants
blameprompt hackathon-report --start "2026-02-26T09:00:00Z" --end "2026-02-26T21:00:00Z"

Generates an integrity report with timeline, code attribution, and anomaly detection.

Sharing & interop

blameprompt push                    # push notes to remote
blameprompt pull                    # fetch notes from remote
blameprompt cache sync              # build local SQLite cache
blameprompt export-agent-trace      # export to Agent Trace v0.1.0 format
blameprompt import-agent-trace      # display Agent Trace record
blameprompt github-comment          # post AI attribution as PR comment

Setup & diagnostics

blameprompt init --global           # global setup (hooks, git template, 15 agents)
blameprompt init                    # setup in current repo only
blameprompt install-git-wrap        # transparent git wrapper (auto-attach on commit)
blameprompt doctor                  # diagnose installation issues
blameprompt update                  # self-update
blameprompt uninstall               # remove hooks, keep receipt history
blameprompt uninstall --purge       # remove everything including Git Notes

What gets captured

Every AI receipt includes: provider, model, user, timestamp, session ID, prompt & response summaries, files changed (with line ranges), token usage (input, output, cache read, cache creation), real token-based cost, tools used, MCP servers called, agents spawned, conversation chain of thought, prompt quality score (4-dimension: clarity, actionability, context, efficiency), prompt category, acceptance rate, and parent receipt links.

Cost tracking uses actual API token data — cache reads at 90% discount, cache creation at 25% surcharge. Pricing for Claude, GPT-4o/4.1/o1/o3, Gemini 2.5, Codex, and more.

Privacy & data

Everything is local-first. Nothing leaves your machine unless you push or sync.

WhatWhere
AI receiptsrefs/notes/blameprompt (inside .git)
Staging.blameprompt/staging.json (gitignored)
Credentials~/.blameprompt/credentials
Cache~/.blameprompt/prompts.db
Config.blamepromptrc or ~/.blamepromptrc

Zero telemetry. Zero tracking. Built-in redaction engine strips secrets before storage. You choose what to sync to your public profile.

Enterprise

BlamePrompt Enterprise provides team-level AI code observability, compliance reporting, and adoption benchmarking. Visit blameprompt.com/enterprise.

License

MIT