diff --git a/ADOPTERS.md b/.github/ADOPTERS.md
similarity index 100%
rename from ADOPTERS.md
rename to .github/ADOPTERS.md
diff --git a/.github/pages-stub/index.html b/.github/pages-stub/index.html
index 39deba000..7bf034d1c 100644
--- a/.github/pages-stub/index.html
+++ b/.github/pages-stub/index.html
@@ -12,7 +12,7 @@
     -->
     <meta
       name="description"
-      content="Coder Eval is Playwright for coding agents: an open-source, agent-agnostic framework for evaluating and benchmarking AI coding agents and their skills — sandboxed runs of Claude Code, Codex, Antigravity (Gemini), OpenCode, or Pi against declarative YAML tasks, with weighted scoring and CI gates. Documentation: coder-eval.com/docs"
+      content="Coder Eval is Playwright for coding agents: test that your skills, MCP servers, and CLIs actually work when an agent uses them — sandboxed YAML evals for Claude Code, Codex, Antigravity (Gemini), OpenCode and Pi, with weighted scoring, A/B experiments and CI gates. Documentation: coder-eval.com/docs"
     />
 
     <!--
diff --git a/.github/social-preview/README.md b/.github/social-preview/README.md
new file mode 100644
index 000000000..0820cf857
--- /dev/null
+++ b/.github/social-preview/README.md
@@ -0,0 +1,55 @@
+# Social preview image
+
+`social-preview.png` (1280×640) is the repository's Open Graph card. It is what
+Hacker News, X, Reddit, Slack, and LinkedIn show when someone shares the repo link.
+
+`social-preview.html` is the source. The logo is inlined from
+`.github/pages-stub/index.html`, so the file is self-contained — no external
+stylesheets, fonts, or images.
+
+## Regenerate
+
+```bash
+"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \
+  --headless --disable-gpu --hide-scrollbars \
+  --force-device-scale-factor=2 --window-size=1280,640 \
+  --screenshot=out@2x.png \
+  "file://$PWD/.github/social-preview/social-preview.html"
+
+# downsample the 2× capture to the 1280×640 GitHub expects
+python3 -c "from PIL import Image; \
+  Image.open('out@2x.png').resize((1280,640), Image.LANCZOS) \
+       .save('.github/social-preview/social-preview.png', optimize=True)"
+```
+
+## Publish
+
+GitHub has **no API** for the social preview image. Upload it by hand:
+
+**Settings → General → Social preview → Upload an image**
+<https://github.com/UiPath/coder_eval/settings>
+
+## What the card says
+
+| Slot | Copy |
+| --- | --- |
+| Repo | github.com/**UiPath/coder_eval** |
+| Headline | Playwright for coding agents. |
+| Positioning | Test that your **skills**, **MCP servers**, and **CLIs** actually work when an agent uses them. |
+| Where it runs | Run it from the **command line**, in a **sandbox**, as an **A/B experiment**, or as a **CI gate**. |
+| Capabilities | YAML suites · Activation checks · Weighted scoring · Cost telemetry |
+| Footer | Apache-2.0 · Python 3.13+ · Claude Code · Codex · Gemini · OpenCode · Pi |
+
+The terminal panel deliberately shows `runs/latest`, not a dated run directory —
+a timestamp in the image tells every future reader when the card was made.
+
+The positioning and capability lines mirror the repo's GitHub description. Change
+one and change the other, or a shared link and the repo page will say two different
+things.
+
+## Keep it honest
+
+The card names every supported harness. When a new `agent.type` ships, this image
+is a surface that has to change with it — the same rule CE047 enforces on the
+README, `docs/index.md`, `docs/comparison.md`, `docs/llms.txt`, `mkdocs.yml`, the
+Pages stub, and `pyproject.toml`. It is not lint-enforced, because it is a PNG.
diff --git a/.github/social-preview/social-preview.html b/.github/social-preview/social-preview.html
new file mode 100644
index 000000000..cf210a211
--- /dev/null
+++ b/.github/social-preview/social-preview.html
@@ -0,0 +1,149 @@
+<!doctype html>
+<html><head><meta charset="utf-8">
+<style>
+  * { margin:0; padding:0; box-sizing:border-box; }
+  html, body { width:1280px; height:640px; }
+  body {
+    background:#0B1826;
+    font-family:-apple-system, "SF Pro Text", "Helvetica Neue", Helvetica, Arial, sans-serif;
+    color:#E8EEF5;
+    overflow:hidden;
+    position:relative;
+    -webkit-font-smoothing:antialiased;
+  }
+  .grid {
+    position:absolute; inset:0;
+    background-image:
+      linear-gradient(to right, #0E1C2A 1px, transparent 1px),
+      linear-gradient(to bottom, #0E1C2A 1px, transparent 1px);
+    background-size:64px 64px;
+  }
+  .glow {
+    position:absolute; right:-180px; bottom:-240px;
+    width:820px; height:740px; border-radius:50%;
+    background:radial-gradient(circle, rgba(250,70,22,0.15) 0%, rgba(250,70,22,0.035) 46%, transparent 70%);
+  }
+  .bar { position:absolute; left:0; right:0; bottom:0; height:10px; background:#FA4616; }
+
+  .wrap { position:relative; display:flex; height:100%; padding:62px 68px; gap:40px; }
+  .left  { width:558px; display:flex; flex-direction:column; }
+  .right { flex:1; display:flex; align-items:center; }
+
+  .kicker {
+    font-family:"SF Mono", Menlo, monospace;
+    font-size:15.5px; letter-spacing:0.01em; font-weight:500;
+    color:#6E8399;
+  }
+  .kicker b { color:#FA4616; font-weight:600; }
+  .logo { height:42px; width:134px; color:#FFFFFF; margin-top:24px; display:block; align-self:flex-start; }
+
+  h1 {
+    margin-top:22px;
+    font-size:45px; line-height:1.08; font-weight:800; letter-spacing:-0.025em;
+    color:#FFFFFF;
+  }
+  h1 .accent { color:#FA4616; }
+
+  .sub { margin-top:16px; font-size:19.5px; line-height:1.45; color:#A7B8CA; }
+  .sub b { color:#FFFFFF; font-weight:600; }
+
+  .runline {
+    margin-top:13px; font-size:18px; line-height:1.45; color:#7E93A9;
+  }
+  .runline b { color:#C6D5E3; font-weight:600; }
+
+  .chips { margin-top:18px; display:flex; flex-wrap:nowrap; gap:8px; }
+  .chip {
+    border:1px solid #1E3C5E; border-radius:999px;
+    padding:6px 12px; font-size:13px; color:#8FA4B9;
+    background:rgba(16,34,55,0.6); white-space:nowrap;
+  }
+
+  .install {
+    margin-top:22px; align-self:flex-start;
+    border:1px solid #1C3A5C; border-radius:9px;
+    padding:13px 20px;
+    font-family:"SF Mono", Menlo, monospace; font-size:18px;
+    background:rgba(13,30,49,0.55);
+  }
+  .install .p { color:#4D6party; }
+  .install .dollar { color:#5C7governance; }
+
+  .foot {
+    position:absolute; left:68px; bottom:54px;
+    font-size:17px; color:#7E93A9; letter-spacing:0.005em;
+  }
+  .foot b { color:#BACBDC; font-weight:600; }
+
+  /* terminal panel */
+  .term {
+    width:100%; background:#091420; border:1px solid #16293D; border-radius:12px;
+    box-shadow:0 26px 70px rgba(0,0,0,0.42);
+    overflow:hidden;
+  }
+  .term-head {
+    display:flex; align-items:center; gap:10px;
+    padding:13px 16px; border-bottom:1px solid #16293D;
+  }
+  .dot { width:11px; height:11px; border-radius:50%; }
+  .term-title {
+    margin-left:6px; background:#122437; border-radius:6px;
+    padding:4px 11px; font-family:"SF Mono", Menlo, monospace;
+    font-size:13px; color:#8A95A2;
+  }
+  .term-body {
+    padding:20px 22px 24px;
+    font-family:"SF Mono", Menlo, monospace; font-size:14.5px; line-height:1.85;
+    color:#C3CEDA; white-space:pre;
+  }
+  .c-dim { color:#6B7B8C; }
+  .c-org { color:#FA6A3C; }
+  .c-grn { color:#3FCF6E; }
+  .c-wht { color:#FFFFFF; font-weight:600; }
+  .c-blu { color:#5BA7F7; }
+  .barfill { color:#3FCF6E; }
+</style></head>
+<body>
+  <div class="grid"></div>
+  <div class="glow"></div>
+
+  <div class="wrap">
+    <div class="left">
+      <div class="kicker">github.com/<b>UiPath/coder_eval</b></div>
+      <svg class="logo" aria-hidden="true" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1400 440" fill="none"><g transform="translate(-75,-36)"><path d="M224 176 L216 176 A56 56 0 0 0 160 232 L160 280 A56 56 0 0 0 216 336 L224 336" fill="none" stroke="#FA4616" stroke-width="28" stroke-linecap="round" stroke-linejoin="round"/><rect x="200" y="210" width="78" height="18" rx="9" fill="currentColor"/><rect x="200" y="247" width="52" height="18" rx="9" fill="currentColor"/><rect x="200" y="284" width="104" height="18" rx="9" fill="currentColor"/></g><g fill="currentColor"><g transform="translate(270.00,278.00) scale(0.19000,-0.19000)"><path d="M308 -10Q237 -10 185.0 16.5Q133 43 104.5 91.5Q76 140 76 206V524Q76 590 104.5 638.5Q133 687 185.0 713.5Q237 740 308 740Q378 740 430.0 713.5Q482 687 511.0 638.5Q540 590 540 524H414Q414 575 386.5 602.5Q359 630 308 630Q257 630 229.0 602.5Q201 575 201 524V206Q201 155 229.0 127.5Q257 100 308 100Q359 100 386.5 127.5Q414 155 414 206H540Q540 140 511.0 91.5Q482 43 430.0 16.5Q378 -10 308 -10Z"/></g><g transform="translate(380.00,278.00) scale(0.19000,-0.19000)"><path d="M300 -9Q194 -9 132.0 50.5Q70 110 70 211V339Q70 441 132.0 500.0Q194 559 300 559Q406 559 468.0 500.0Q530 441 530 339V211Q530 110 468.0 50.5Q406 -9 300 -9ZM300 100Q350 100 377.5 127.5Q405 155 405 207V343Q405 395 377.5 422.5Q350 450 300 450Q250 450 222.5 422.5Q195 395 195 343V207Q195 155 222.5 127.5Q250 100 300 100Z"/></g><g transform="translate(490.00,278.00) scale(0.19000,-0.19000)"><path d="M250 -10Q170 -10 120.5 47.0Q71 104 71 200V349Q71 446 120.0 503.0Q169 560 250 560Q307 560 346.5 531.0Q386 502 402 450L398 576V730H523V0H401V100Q386 48 347.0 19.0Q308 -10 250 -10ZM297 98Q345 98 371.5 127.0Q398 156 398 209V341Q398 394 371.5 423.0Q345 452 297 452Q249 452 222.5 426.0Q196 400 196 346V204Q196 150 222.5 124.0Q249 98 297 98Z"/></g><g transform="translate(600.00,278.00) scale(0.19000,-0.19000)"><path d="M301 -10Q231 -10 179.0 17.0Q127 44 98.5 93.5Q70 143 70 210V340Q70 407 98.5 456.5Q127 506 179.0 533.0Q231 560 301 560Q370 560 421.5 533.0Q473 506 501.5 457.5Q530 409 530 344V244H190V206Q190 150 219.0 121.5Q248 93 301 93Q340 93 367.5 106.5Q395 120 400 146H524Q511 75 450.0 32.5Q389 -10 301 -10ZM190 344V325L410 326V345Q410 402 382.0 433.0Q354 464 301 464Q246 464 218.0 432.5Q190 401 190 344Z"/></g><g transform="translate(710.00,278.00) scale(0.19000,-0.19000)"><path d="M97 0V550H219V443Q229 497 268.5 528.5Q308 560 371 560Q455 560 502.0 506.0Q549 452 549 355V303H424V349Q424 399 398.0 425.5Q372 452 324 452Q274 452 248.0 422.0Q222 392 222 337V0Z"/></g></g><g fill="#FA4616"><g transform="translate(890.00,278.00) scale(0.19000,-0.19000)"><path d="M88 0V730H526V620H211V430H491V324H211V110H526V0Z"/></g><g transform="translate(1000.00,278.00) scale(0.19000,-0.19000)"><path d="M221 0 36 550H167L273 216Q283 186 290.5 155.5Q298 125 301 103Q305 125 311.5 155.5Q318 186 327 215L432 550H564L380 0Z"/></g><g transform="translate(1110.00,278.00) scale(0.19000,-0.19000)"><path d="M239 -10Q155 -10 106.5 37.5Q58 85 58 162Q58 242 113.0 289.5Q168 337 264 337H398V376Q398 459 301 459Q257 459 231.0 442.0Q205 425 202 394H82Q86 468 144.5 514.0Q203 560 302 560Q407 560 465.0 511.0Q523 462 523 373V0H401V103Q394 50 351.0 20.0Q308 -10 239 -10ZM279 92Q334 92 366.0 119.0Q398 146 398 191V256H270Q229 256 205.0 233.0Q181 210 181 174Q181 137 206.5 114.5Q232 92 279 92Z"/></g><g transform="translate(1220.00,278.00) scale(0.19000,-0.19000)"><path d="M389 0Q307 0 257.5 49.0Q208 98 208 178V617H27V730H333V181Q333 150 350.5 131.5Q368 113 398 113H559V0Z"/></g></g></svg>
+      <h1>Playwright for<br>coding <span class="accent">agents</span>.</h1>
+      <div class="sub">Test that your <b>skills</b>, <b>MCP servers</b>, and <b>CLIs</b><br>actually work when an agent uses them.</div>
+      <div class="runline">Run it from the <b>command line</b>, in a <b>sandbox</b>,<br>as an <b>A/B experiment</b>, or as a <b>CI gate</b>.</div>
+      <div class="chips">
+        <span class="chip">YAML suites</span>
+        <span class="chip">Activation checks</span>
+        <span class="chip">Weighted scoring</span>
+        <span class="chip">Cost telemetry</span>
+      </div>
+      <div class="install"><span class="c-dim">$</span> uv tool install coder-eval</div>
+    </div>
+
+    <div class="right">
+      <div class="term">
+        <div class="term-head">
+          <span class="dot" style="background:#FF5F57"></span>
+          <span class="dot" style="background:#FEBC2E"></span>
+          <span class="dot" style="background:#28C840"></span>
+          <span class="term-title">coder-eval run</span>
+        </div>
+        <div class="term-body"><span class="c-dim">$</span> coder-eval run tasks/*.yaml
+<span class="c-wht">Run directory:</span> runs/latest
+<span class="c-wht">Running</span> <span class="c-org">4</span> <span class="c-wht">task(s)</span>  <span class="c-dim">Agent: claude-code</span>
+<span class="c-blu">[INFO]</span> skill_triggered: <span class="c-grn">PASS</span>   score 1.000
+<span class="c-blu">[INFO]</span> file_exists:     <span class="c-grn">PASS</span>   score 1.000
+Tasks: 100%|<span class="barfill">████████████████</span>| 4/4
+<span class="c-grn">Run complete:</span>  <span class="c-dim">$0.14 · 21.6s</span>
+<span class="c-wht">Results:</span> <span class="c-grn">4/4</span> succeeded   score <span class="c-org">1.000</span></div>
+      </div>
+    </div>
+  </div>
+
+  <div class="foot">Apache-2.0 &middot; Python 3.13+ &middot; <b>Claude Code &middot; Codex &middot; Gemini &middot; OpenCode &middot; Pi</b></div>
+
+  <div class="bar"></div>
+</body></html>
diff --git a/.github/social-preview/social-preview.png b/.github/social-preview/social-preview.png
new file mode 100644
index 000000000..baeb9d9c2
Binary files /dev/null and b/.github/social-preview/social-preview.png differ
diff --git a/README.md b/README.md
index e0b78755a..abc6a788f 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# Coder Eval — evaluate and benchmark AI coding agents and their skills
+# Coder Eval — Playwright for coding agents — test that your skills, MCP servers, and CLIs actually work when an agent uses them.
 
 [![PyPI](https://img.shields.io/pypi/v/coder-eval.svg)](https://pypi.org/project/coder-eval/)
 [![GitHub Marketplace](https://img.shields.io/badge/marketplace-coder__eval-2ea44f.svg)](https://github.com/marketplace/actions/coder_eval)
@@ -9,40 +9,41 @@
 [![CI](https://github.com/UiPath/coder_eval/actions/workflows/pr-checks.yml/badge.svg)](https://github.com/UiPath/coder_eval/actions/workflows/pr-checks.yml)
 
 <p align="center">
-  <strong>Playwright for coding agents</strong> — one declarative test file, any agent
-  runtime, a real sandbox, and a pass/fail gate in CI.
+  Run it from the <strong>command line</strong>, in a <strong>sandbox</strong>, as an
+  <strong>A/B experiment</strong>, or as a <strong>CI gate</strong>.
 </p>
 
 **Coder Eval** (`pip install coder-eval` / `uv tool install coder-eval`) is an
-open-source, **agent-agnostic** framework for **evaluating and benchmarking AI coding
-agents and their skills** — built for benchmark authors, CLI builders, and skill
-builders — with sandboxing, reproducibility, and data-driven analysis. It runs a real
-agent — **Claude Code**, **OpenAI Codex**, **Google Antigravity (Gemini)**,
-**OpenCode**, or **Pi** — in a sandbox against declarative YAML tasks, then scores the files and
-commands it actually produced. Changing harness is one field (`agent.type`); the
-tasks, criteria, scoring, telemetry, and reports stay the same.
-
-Reach for it when you want to **benchmark agents on your own domain tasks**,
-**test whether a skill triggers** in the agent you ship for, **A/B-test Claude Code
-vs. Codex vs. Gemini vs. OpenCode vs. Pi** (or model vs. model, prompt vs. prompt), or
-**gate CI on coding-agent quality**. It is **not a fixed leaderboard**: unlike
-SWE-bench or SkillsBench, which rank models on a shared task set, you bring the tasks
-and you bring the scoring — weighted 0.0–1.0 criteria, a `skill_triggered` activation
-check, an A/B experiment layer, and per-tool cost telemetry, over whatever work *you*
-care about. See [How it compares](https://coder-eval.com/docs/comparison).
-📚 **Full docs:** **[coder-eval.com/docs](https://coder-eval.com/docs)**.
+open-source, **agent-agnostic** framework for evaluating and benchmarking AI coding
+agents and their skills — built for benchmark authors, CLI builders, and skill
+builders. It runs a real agent — **Claude Code**, **OpenAI Codex**, **Google
+Antigravity (Gemini)**, **OpenCode**, or **Pi** — in a sandbox against declarative
+YAML tasks, then scores the files and commands it actually produced. Changing harness
+is one field (`agent.type`). It is **not a fixed leaderboard**: unlike SWE-bench or
+SkillsBench, you bring the tasks and you bring the scoring — weighted 0.0–1.0
+criteria, a `skill_triggered` activation check, an A/B experiment layer, and per-tool
+cost telemetry. See [How it compares](https://coder-eval.com/docs/comparison).
+
+📚 **Full docs:** **[coder-eval.com/docs](https://coder-eval.com/docs)**
 
 <p align="center">
   <img src="docs/assets/hero.gif" alt="Coder Eval running the hello_date task: a sandboxed agent writes and runs a script from a YAML task, then the scored result is browsed in evalboard" width="100%">
 </p>
 
-- **Declarative YAML tasks** with pinned dependencies and clear success criteria
-- **Sandboxed execution** in isolated environments with resource limits
-- **Weighted, continuous scoring** (0.0–1.0) with fractional credit and thresholds
-- **Many criterion types** — from file checks to code similarity and LLM-graded rubrics
-- **Agent-agnostic by design** — Claude Code, OpenAI Codex, Antigravity (Gemini), OpenCode, and Pi today; add your own harness through the plugin SPI
-- **Experiment layer** — A/B agent configs (models, tools, prompts) side-by-side
+- **Declarative YAML tasks** — a prompt, a sandbox, and weighted 0.0–1.0 success criteria in one file
+- **Agent-agnostic by design** — Claude Code, OpenAI Codex, Antigravity (Gemini), OpenCode, and Pi today; swap harness with one field, or add your own through the plugin SPI
+- **Sandboxed execution** — isolated environments with resource limits, on a tempdir or a container
+- **Many criterion types** — from file and command checks to code similarity and LLM-graded rubrics
+- **Skill-activation testing** — verify an agent actually engages a target skill (`skill_triggered`), on whichever harness your users run
+- **A/B experiments** — model vs. model, tool-on vs. tool-off, prompt vs. prompt, side by side
 - **Full telemetry** — every tool call, token counts, and cost, with real-time streaming
+- **CI gate** — run the suite in GitHub Actions and fail the build on a regression
+- **Bring your own dataset** — fan one task out over many rows for larger benchmark suites
+
+> **Keeping skills fresh?** Run Coder Eval as a scheduled GitHub Actions job so your
+> skills are continuously re-evaluated against the latest model — a skill that quietly
+> stops triggering surfaces as a failing criterion before your users hit it. See
+> **[Tutorial 02 — Running Coder Eval in CI](docs/tutorials/02-ci-pipeline.md)**.
 
 ## Watch the intro
 
@@ -59,26 +60,12 @@ care about. See [How it compares](https://coder-eval.com/docs/comparison).
 ▶ Also on YouTube: **[Coder Eval: UiPath open-source framework to test AI Coding Agents](https://www.youtube.com/watch?v=Iyq-5m1CnuI)**
 — what the framework does, and how a run works end to end.
 
-## What you can do with it
-
-- **Benchmark coding agents** — score an agent across a suite of tasks with weighted scoring and pass/fail thresholds
-- **Compare models & configs** — A/B-test Claude vs. Codex vs. Gemini vs. OpenCode vs. Pi, model vs. model, tool-on vs. tool-off, prompt vs. prompt
-- **Evaluate skills** — verify an agent actually engages a target skill (`skill_triggered`) and score skill-driven suites (SkillsBench-style), on whichever harness your users run
-- **Keep skills up to date in CI** — re-validate your skills on every change or on a schedule; catch silent regressions when models, prompts, or the skills themselves drift
-- **Gate CI on agent quality** — run the suite in GitHub Actions and fail the build on regressions
-- **Bring your own dataset** — fan one task out over many rows for larger benchmark suites
-
-> **Keeping skills fresh?** Run Coder Eval as a scheduled GitHub Actions job so your
-> skills are continuously re-evaluated against the latest model — a skill that quietly
-> stops triggering surfaces as a failing criterion before your users hit it. See
-> **[Tutorial 02 — Running Coder Eval in CI](docs/tutorials/02-ci-pipeline.md)**.
-
 ## Quick Start
 
 **Prerequisites:** Python 3.13+, [uv](https://docs.astral.sh/uv/) 0.8+, and **the
 runtime of at least one coding agent** — plus that agent's own model credentials.
-Pick the agent you want to evaluate; two of the four ship with a Coder Eval extra,
-the other two are separate CLIs you install yourself:
+Pick the agent you want to evaluate; two of the five ship with a Coder Eval extra,
+the other three are separate CLIs you install yourself:
 
 | Agent | `agent.type` | Runtime | Guide |
 | --- | --- | --- | --- |
@@ -86,6 +73,7 @@ the other two are separate CLIs you install yourself:
 | OpenAI Codex | `codex` | `uv sync --extra codex` — the extra ships the Codex SDK + CLI | [Codex](docs/agents/CODEX.md) |
 | Google Antigravity (Gemini) | `antigravity` | `uv sync --extra antigravity` — the extra ships the harness binary | [Antigravity](docs/agents/ANTIGRAVITY.md) |
 | OpenCode (open-weight models) | `opencode` | `npm install -g opencode-ai` — separate CLI | [OpenCode](docs/agents/OPENCODE.md) |
+| Pi (open-weight models) | `pi` | `npm install -g @earendil-works/pi-coding-agent` — separate CLI | [Pi](docs/agents/PI.md) |
 
 The examples below use the default `claude-code` agent. Developed on macOS; CI runs on
 Linux.
@@ -131,6 +119,9 @@ live in this repo — clone it or point the CLI at your own task files.) See
 [Tutorial 02 — Running Coder Eval in CI](docs/tutorials/02-ci-pipeline.md) for
 the full setup.
 
+> ⭐ **If this saved you an afternoon, star the repo** — it is how other people
+> evaluating coding agents find it.
+
 ## Use inside Claude Code
 
 Coder Eval evaluates any of the supported agents, and it also ships an authoring
@@ -312,9 +303,9 @@ success_criteria:
     description: "Script must execute successfully"
 ```
 
-`agent.type` is the only harness-specific line: swap it for `codex`, `antigravity`, or
-`opencode` — or override it per run with `coder-eval run … -D agent.type=opencode` — and
-the same criteria score the same way. Tasks can omit the `agent` section entirely —
+`agent.type` is the only harness-specific line: swap it for `codex`, `antigravity`,
+`opencode`, or `pi` — or override it per run with `coder-eval run … -D agent.type=opencode`
+— and the same criteria score the same way. Tasks can omit the `agent` section entirely —
 defaults resolve from the experiment layer (`experiments/default.yaml`). For the full
 schema and every criterion type, see the
 [Task Definition Guide](docs/TASK_DEFINITION_GUIDE.md); for what each `run_limits` field
@@ -366,4 +357,4 @@ Built with [Pydantic](https://pydantic.dev/), [Typer](https://typer.tiangolo.com
 and [Rich](https://rich.readthedocs.io/), on top of the harnesses it drives — the
 [Claude Agent SDK](https://github.com/anthropics/claude-agent-sdk), the
 [Codex SDK](https://github.com/openai/codex), [Google Antigravity](https://antigravity.google/),
-and [OpenCode](https://opencode.ai).
+[OpenCode](https://opencode.ai), and [Pi](https://pi.dev/).
diff --git a/action.yml b/action.yml
index 6f2fe764d..1f00d628d 100644
--- a/action.yml
+++ b/action.yml
@@ -3,7 +3,7 @@
 # Display only: consumers reference the action by repo path.
 name: coder_eval
 author: UiPath
-description: Install a pinned coder-eval and run evaluation tasks as a CI gate, with JUnit XML output.
+description: Gate CI on coding-agent quality — run sandboxed coder-eval suites and fail the build on a regression, with JUnit XML output.
 branding:
   icon: check-circle
   color: orange
diff --git a/comparison.md b/comparison.md
deleted file mode 100644
index dd315b9bf..000000000
--- a/comparison.md
+++ /dev/null
@@ -1,171 +0,0 @@
-# Closing the grading-material leak: two approaches compared
-
-> **Scope note.** This is a design-decision document about *sandbox isolation*. It is unrelated to
-> `docs/comparison.md`, which compares coder_eval to other eval frameworks (SWE-bench, Harbor, …).
-
-- **Approach A** — [PR #88](https://github.com/UiPath/coder_eval/pull/88),
-  `fix/docker-isolation-copy-prune`: **COPY/PRUNE + GRADE-OUTSIDE**. Open, draft, 8 commits,
-  +5,754 / −214 across 35 files. Implemented and verified.
-- **Approach B** — **Declarative trust split + DAC permissions**: `input_dir` / `reference_dir` on
-  `TaskDefinition`, `$INPUT_DIR` / `$REFERENCE_DIR` substitution, agent runs as an unprivileged uid,
-  `reference_dir` is root-owned `0700`, `task.yaml` itself lives in `reference_dir`. A design, not
-  yet built.
-
----
-
-## The problem
-
-When coder_eval grades an agent, the grading material sits on the same filesystem the agent is
-working in — **including under `--driver docker`, which is the mode users assume is safe**.
-
-On `main`, an agent in the container can reach all of the following:
-
-| Surface | Where it lives today |
-|---|---|
-| Full `success_criteria` | `/work/input/task.yaml`, staged verbatim (`docker_runner.py:605-620`) |
-| Raw task YAML text | `context.json.source_yaml` |
-| Reference solution | `reference.file` / `reference.directory`, auto-mounted (`:1253-1254`) |
-| Task-dir graders (`check_*.py`, `RESOLUTION.md`, reference agents) | host task dir bind-mounted at its own host path (`:1176`) |
-| Signposts pointing at all of the above | `TASK_DIR` / `SKILLS_REPO_PATH` in the agent's env |
-
-The agent and the harness run as the **same user** in the container — both root, since neither
-`docker/Dockerfile` nor `docker/Dockerfile.runtime` sets `USER` and `_build_argv` passes no
-`--user`. So there is no permission barrier of any kind: a single `grep -r` reaches the answer key.
-
-**This is measured, not theoretical.** PR #88 reports a nightly audit in which **~2.4% of replicates
-read the suite instead of solving the task**, and an adversarial "check `/work`" prompt drives that
-to **~100%**.
-
-A separate but adjacent defect: `~/.uipath` (including `.auth`) was mounted **read-write from the
-host original**, so an agent could overwrite operator credentials.
-
-The root cause is structural: **the harness lives inside the sandbox it is supposed to be isolating
-from.** The in-container process runs the agent *and* grades it (`run_task_internal_command.py:190-209`),
-so the rubric has to be shipped into the container at `t=0`.
-
----
-
-## Approach A — COPY/PRUNE + GRADE-OUTSIDE (PR #88)
-
-**Closes the leak by absence, not by permission.** The container never contains grading material.
-
-Three coordinated moves:
-
-1. **COPY/PRUNE the agent's inputs.** The staged `task.yaml` is criteria-stripped via
-   `agent_safe_dump()` (`success_criteria: []`, `reference: null`) and `context.json.source_yaml` is
-   nulled. Plugins are projected to a *sanitized bundle copy* (`project_plugin_for_agent`, allowlist
-   `PLUGIN_AGENT_ALLOWED_SUBDIRS` = `skills`/`commands`/`agents`/`hooks`/`.claude-plugin`) mounted
-   `:ro` at `/work/skills`. The raw checkout, the reference, and the host task dir are **not mounted
-   at all**. A **grader-dir overlap guard** hard-errors any `template_sources` / `system_prompt_file`
-   / `extra_mounts` whose source contains the task dir.
-2. **GRADE OUTSIDE.** The container runs the agent only. After it exits, the host grades the
-   copied-out artifacts via `regrade_on_host`, using the full unstripped `TaskDefinition` it still
-   holds, with `TASK_DIR` pointing at the real host task dir. Re-grade is restricted to an allowlist
-   of final statuses (`SUCCESS` / `FAILURE` / `MAX_TURNS_EXHAUSTED`).
-3. **Copy-then-mount host state.** `~/.uipath` and `~/.claude` are forwarded as throwaway rw
-   **copies**; `~/.claude/jobs` (operator session timelines) is denylisted.
-
-Plus **harness-outside**: `pre_run`/`post_run` invoke helper scripts under the skills-repo `tests/`
-tree, which is no longer mounted. `post_run` becomes host-only; `pre_run` gains a per-command
-`runs_in: host|agent` (default `host`, seeding the workspace through a `:ro` `/work/seed` staging
-mount). The PR body flags this as a known blocker affecting **243 of 1,147 downstream tasks (~21%)**;
-the branch implements the fix, with only the ~7 `uv sync` tasks needing `runs_in: agent`.
-
-**Verification shipped with it:** Detector A (host-unchanged — daemon-less mount-shape proxy in CI
-plus an `-m live` byte-and-metadata-identical real run), Detector B (zero-grading-material in the
-agent mount view, with a positive control so "staged nothing" can't pass vacuously), a baked-image
-answer scan, and live confirmation that an adversarial "read your grader" task fails on gpt / claude
-/ gemini / kimi.
-
-**Stated costs:** early stop is **disabled under docker** (criteria are gone, so the in-container
-watcher can't arm — a warning is logged, verdict unaffected), and the design doc states plainly that
-**allowlist-by-absence has no DAC backstop** — if the prune boundary is wrong, nothing else stops the
-read.
-
-## Approach B — Declarative trust split + DAC permissions
-
-**Closes the leak by permission, with an author-declared boundary.**
-
-- `TaskDefinition` gains `input_dir` (agent-visible fixtures) and `reference_dir` (answer key),
-  substituted as `$INPUT_DIR` / `$REFERENCE_DIR` inside the task file.
-- `task.yaml` itself lives in `reference_dir`, so the rubric is protected by the same single rule as
-  everything else — no special case.
-- In the container, the agent runs as an unprivileged uid; `reference_dir` is root-owned `0700`;
-  `input_dir` and the workspace are chowned to the agent uid.
-- Effectively this **splits the existing `$TASK_DIR` token into its two halves by trust level** —
-  which is precisely why the leak exists today, since one token points at a directory mixing both.
-
-Refinements this needs to be sound:
-
-- **Scope-aware substitution.** `$REFERENCE_DIR` must be a hard validation error in agent-facing
-  fields (`initial_prompt`, `system_prompt`, mutations) — resolving it there advertises the path and
-  leaks it into logs and reports. `$INPUT_DIR` substitutes anywhere.
-- **Copy, don't mount, `reference_dir`.** The host bind mount is `:ro`, so `chmod` fails `EROFS`;
-  making it writable would mutate the user's real repo permissions with no crash-safe restore. Stage
-  a container-local root-owned copy — which also lets the symmetric same-host-path mount disappear.
-- **Static, not scoped, permissions.** A `with block_access(): …` context manager that flips modes
-  on entry and restores on exit is not crash-safe (SIGKILL between enter and exit leaves them
-  flipped). With two identities, the perms are simply set once at stage time and never restored.
-- **Dataset files default to `reference_dir`** — for activation suites the JSONL rows carry
-  `expected_skill`, which *is* the answer key.
-- The `~/.claude` credential copy must stay agent-writable, so `0700`-root cannot be applied
-  blanket — it is per-path.
-
----
-
-## Pros and cons
-
-| | Pros | Cons |
-|---|---|---|
-| **A — COPY/PRUNE + GRADE-OUTSIDE** (PR #88) | Closes the leak **even against a root agent** — no identity work needed. Fails safe: a file that isn't there can't be read. Also closes surfaces B doesn't touch: host-credential overwrite (`~/.uipath` rw), `~/.claude/jobs` exposure, env signposts, baked-image content. **Built, tested, and empirically verified** (detectors in CI + live adversarial probe across four models). Requires **no per-task authoring changes** for the leak fix itself. Removes the harness from the sandbox — the structural root cause. | **Early stop disabled under docker** — a shipped feature silently degrades on one driver. Forced `pre_run`/`post_run` host-side, touching **~21% of downstream tasks** and adding a new `runs_in` concept. **No DAC backstop** (the PR says so): the prune allowlist is "a coder_eval-side guess about what is answer-free", so a plugin putting answers in `skills/` defeats it. Detector B catches known sentinels, not an unknown golden file. Large blast radius: **+5,754 LOC / 35 files**, restructuring `orchestrator.py`. Nothing protects `tempdir`. |
-| **B — Trust split + DAC permissions** | Supplies exactly the **DAC backstop A lacks** — a second, independent barrier. The boundary is **author-declared**, not framework-guessed, replacing A's residual risk #1 with an explicit contract. **Preserves early stop** (criteria stay in container memory, which the agent can't read) and **keeps `pre_run`/`post_run` in-container** (they run as root and can still read `reference_dir`) — no `runs_in`, no 243-task migration. Much smaller runtime change: a uid on spawn plus chowns at setup. Conceptually simple: one rule, "`reference_dir` is harness-only". | **Depends entirely on identity separation that doesn't exist yet** — both processes are root today, so `chmod 0700` is a *no-op* until the agent runs as a separate uid. Fails **open and silently**: one missed chown, one new mount, and there's no absence to fall back on. Real friction running the agent CLI unprivileged (HOME, npm cache, the `~/.claude` copy must stay writable). Pushes classification onto **1,147 downstream task authors** unless the default is "protected unless declared input". Does **not** address host-credential overwrite, baked-image content, or env signposts. **Also nothing for `tempdir`** (no second identity on the host). Unbuilt and unverified. |
-
-### Head-to-head
-
-| Axis | A — Absence | B — Permissions |
-|---|---|---|
-| Works against a root agent | ✅ | ❌ (requires uid split first) |
-| Failure mode | Fails safe (nothing to read) | Fails open (silently, if a chown is missed) |
-| Boundary defined by | Framework allowlist (a guess) | Task author (a declaration) |
-| Early stop under docker | ❌ disabled | ✅ preserved |
-| `pre_run` / `post_run` | Moved host-side; `runs_in` added; ~21% of tasks affected | Unchanged, in-container |
-| Host credential / `~/.claude/jobs` / baked image | ✅ covered | ❌ out of scope |
-| `tempdir` driver | ❌ | ❌ |
-| Per-task migration | None for the leak fix | `input_dir` / `reference_dir` across the suite |
-| Framework blast radius | +5,754 LOC, 35 files | Smaller runtime change; new schema + token rules |
-| Status | Implemented, CI-gated, live-verified | Design only |
-
----
-
-## Assessment
-
-**These are not competing designs — B is the missing layer under A.**
-
-PR #88's own residual-risk list names its weakest point: *"Allowlist-by-absence has no DAC backstop
-… `PLUGIN_AGENT_ALLOWED_SUBDIRS` is a coder_eval-side guess about what is answer-free."* That is
-precisely the gap Approach B fills, in two independent ways:
-
-1. **As a mechanism** — a uid split gives the second barrier, so a prune-boundary miss is no longer
-   game over.
-2. **As a contract** — and this matters more. `input_dir`/`reference_dir` replaces the framework's
-   guess with an author declaration, which is the durable fix PR #88 already identifies as a
-   cross-repo follow-up ("push the agent-bundle boundary into the skills repo — a manifest declaring
-   the agent-safe surface"). Approach B *is* that manifest, expressed in the task schema.
-
-If only one ships, it should be **A**: it is built, measured against real leak rates, gated by
-detectors in CI, and it closes several surfaces B never touches. B's mechanism half is also blocked
-on identity work (non-root agent, credential-copy ownership) that A doesn't need.
-
-The sequencing that gets the most value:
-
-1. **Land A.** It is the structural fix — it removes the harness from the sandbox.
-2. **Adopt B's declarative half next**, as the durable replacement for the prune allowlist. It is
-   the higher-value part of B and is independent of the permission mechanism.
-3. **Add B's uid split as defense-in-depth** where material must remain in-container.
-4. **Revisit early stop.** A disables it under docker; the leak-free restoration is a host-side
-   watcher over the live event stream — the host already receives every `ToolStartEvent` /
-   `ToolEndEvent` with full `CommandTelemetry` (`streaming/wire.py`), and can already signal the
-   container through the heartbeat channel. Worth first confirming the feature earns its keep: every
-   `stop_early` usage in this repo is a fixture for testing early stop itself.
-5. **State plainly that `tempdir` is not cheat-resistant** under either approach. Neither has a
-   boundary there; it should be documented as a development driver.
diff --git a/docs/index.md b/docs/index.md
index d3fc34ae7..87c9e5e4e 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -1,10 +1,8 @@
 ---
 title: "Evaluate AI coding agents and their skills — Coder Eval"
 description: >-
-  Coder Eval is Playwright for coding agents: an open-source, agent-agnostic
-  framework to evaluate, benchmark, and A/B-test AI coding agents and their skills
-  in a sandbox — declarative YAML tasks, weighted scoring, cost/token telemetry, and
-  CI gates for Claude Code, Codex, Antigravity (Gemini), OpenCode, and Pi.
+  Playwright for coding agents. Test that your skills, MCP servers, and CLIs
+  actually work when an agent uses them — sandboxed, scored, CI-gated.
 ---
 
 # Evaluate AI coding agents and their skills — Coder Eval
diff --git a/docs/llms.txt b/docs/llms.txt
index 9a8adaa68..8c0ed614e 100644
--- a/docs/llms.txt
+++ b/docs/llms.txt
@@ -1,6 +1,8 @@
 # Coder Eval
 
-> Coder Eval (`pip install coder-eval`) is an open-source, agent-agnostic framework
+> Coder Eval (`pip install coder-eval`) is Playwright for coding agents: test that
+> your skills, MCP servers, and CLIs actually work when an agent uses them. It is an
+> open-source, agent-agnostic framework
 > for evaluating, benchmarking, and A/B-testing AI coding agents and their skills
 > (including Claude Code skills) in a sandbox. It uses declarative YAML tasks with
 > weighted, continuous scoring (0.0–1.0), runs real agents (Claude Code, OpenAI
diff --git a/mkdocs.yml b/mkdocs.yml
index d01681d14..c165b7353 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -1,9 +1,7 @@
 site_name: Coder Eval
 site_description: >-
-  Evaluate and benchmark AI coding agents and their skills — a sandboxed,
-  agent-agnostic framework with declarative YAML eval suites for Claude Code,
-  Codex, Antigravity (Gemini), OpenCode and Pi, A/B experiments, weighted scoring,
-  and CI gates.
+  Playwright for coding agents: sandboxed YAML evals for Claude Code, Codex,
+  Gemini, OpenCode and Pi — weighted scoring, A/B experiments, CI gates.
 site_url: https://coder-eval.com/docs/
 repo_url: https://github.com/UiPath/coder_eval
 repo_name: UiPath/coder_eval
diff --git a/pyproject.toml b/pyproject.toml
index 6f6c66a58..563721fe9 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,7 +1,7 @@
 [project]
 name = "coder-eval"
 version = "0.12.1"
-description = "Evaluate, benchmark, and A/B-test AI coding agents (Claude Code, Codex, Gemini/Antigravity, OpenCode, Pi) with sandboxed, reproducible YAML task suites."
+description = "Playwright for coding agents — test that your skills, MCP servers, and CLIs actually work when an agent uses them."
 readme = "README.md"
 license = "Apache-2.0"
 requires-python = ">=3.13"
