chore: rename project to agent-trials, add LICENSE, correct authorship

- plugin/marketplace renamed to agent-trials; author, homepage and repository
  now point to this project rather than the upstream it started from
- LICENSE added (was missing entirely): MIT under Osama's copyright, with the
  upstream MIT notice retained for the derived agent/skill/command/rule and
  hook-script portions
- README leads with trials and token cost instead of fork positioning;
  attribution demoted to a License & Notice section
- OPTIMIZATION.md reframed as historical technique notes

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: osama

.claude-plugin/marketplace.json

@@ -1,43 +1,41 @@
 {
-  "name": "everything-claude-code",
+  "name": "agent-trials",
   "owner": {
-    "name": "Affaan Mustafa",
-    "email": "affaan@example.com"
+    "name": "Osama",
+    "url": "https://github.com/OsamaA140"
   },
   "metadata": {
-    "description": "Token-optimized fork of everything-claude-code - same coverage, a fraction of the context cost"
+    "description": "Agent trials, employee templates, and token-lean Claude Code configuration"
   },
   "plugins": [
     {
-      "name": "everything-claude-code",
+      "name": "agent-trials",
       "source": "./",
-      "description": "Token-lean Claude Code stack: 9 agents, 26 skills, tested zero-token safety hooks - ~1,260 always-on tokens total, 70-80% cheaper per invocation than the upstream originals",
+      "description": "Behavioural trials for Claude Code agents: employee templates, deterministic pass^k grading, zero-token safety hooks, token-lean agents and skills.",
       "author": {
-        "name": "Affaan Mustafa"
+        "name": "Osama"
       },
-      "homepage": "https://github.com/affaan-m/everything-claude-code",
-      "repository": "https://github.com/affaan-m/everything-claude-code",
+      "homepage": "https://github.com/OsamaA140/agent-trials",
+      "repository": "https://github.com/OsamaA140/agent-trials",
       "license": "MIT",
       "keywords": [
         "agents",
-        "skills",
+        "agent-evaluation",
+        "trials",
+        "employee-templates",
         "hooks",
-        "commands",
-        "tdd",
-        "code-review",
         "security",
-        "best-practices"
+        "token-optimization"
       ],
       "category": "workflow",
       "tags": [
         "agents",
-        "skills",
+        "agent-evaluation",
+        "trials",
+        "employee-templates",
         "hooks",
-        "commands",
-        "tdd",
-        "code-review",
         "security",
-        "best-practices"
+        "token-optimization"
       ]
     }
   ]

.claude-plugin/plugin.json

@@ -1,26 +1,26 @@
 {
-  "name": "everything-claude-code",
+  "name": "agent-trials",
   "version": "2.5.1",
-  "description": "Token-lean Claude Code stack: 9 agents, 26 skills, tested zero-token safety hooks (secret-leak + destructive-command blocking) - ~1,260 always-on tokens total, 70-80% cheaper per invocation than the upstream originals",
+  "description": "Behavioural trials for Claude Code agents: hire-ready employee templates, deterministic pass^k grading against fixtures with planted traps, zero-token safety hooks (secret-leak + destructive-command blocking), and a token-lean agent/skill set (~1,260 always-on tokens).",
   "author": {
-    "name": "Affaan Mustafa",
-    "url": "https://x.com/affaanmustafa"
+    "name": "Osama",
+    "url": "https://github.com/OsamaA140"
   },
-  "homepage": "https://github.com/affaan-m/everything-claude-code",
-  "repository": "https://github.com/affaan-m/everything-claude-code",
+  "homepage": "https://github.com/OsamaA140/agent-trials",
+  "repository": "https://github.com/OsamaA140/agent-trials",
   "license": "MIT",
   "keywords": [
     "claude-code",
     "agents",
-    "skills",
+    "agent-evaluation",
+    "trials",
+    "pass-at-k",
+    "employee-templates",
     "hooks",
-    "commands",
-    "rules",
-    "tdd",
-    "code-review",
     "security",
-    "workflow",
-    "automation",
+    "secret-scanning",
+    "token-optimization",
+    "testing",
     "best-practices"
   ],
   "commands": "./commands",

CONTRIBUTING.md

@@ -184,7 +184,7 @@ Then open a PR with:
 
 ## Questions?
 
-Open an issue or reach out on X: [@affaanmustafa](https://x.com/affaanmustafa)
+Open an issue at [OsamaA140/agent-trials](https://github.com/OsamaA140/agent-trials/issues).
 
 ---
 

LICENSE

@@ -0,0 +1,54 @@
+MIT License
+
+Copyright (c) 2026 Osama
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+---
+
+PORTIONS OF THIS SOFTWARE
+
+Some agent, skill, command, and rule definitions, and parts of the
+cross-platform hook scripts, derive from everything-claude-code by
+Affaan Mustafa (https://github.com/affaan-m/ECC), used under the MIT License:
+
+    MIT License
+
+    Copyright (c) 2026 Affaan Mustafa
+
+    Permission is hereby granted, free of charge, to any person obtaining a
+    copy of this software and associated documentation files (the "Software"),
+    to deal in the Software without restriction, including without limitation
+    the rights to use, copy, modify, merge, publish, distribute, sublicense,
+    and/or sell copies of the Software, and to permit persons to whom the
+    Software is furnished to do so, subject to the following conditions:
+
+    The above copyright notice and this permission notice shall be included in
+    all copies or substantial portions of the Software.
+
+    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+    THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+    FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+    DEALINGS IN THE SOFTWARE.
+
+The trial harness, guards library, agent validator, benchmark, employee
+templates, fixtures, and test suite are original works of this project.

OPTIMIZATION.md

@@ -1,6 +1,8 @@
-# Optimization Notes
+# Optimization Notes (historical)
 
-This is a token-optimized rewrite of [affaan-m/everything-claude-code](https://github.com/affaan-m/everything-claude-code) for Osama's own use. Goal: same coverage and behavior, minimum tokens loaded per invocation.
+How the agent, skill, command, and rule set in this project was compressed to its current size. These notes are kept as a record of technique — what was cut and why — and are useful if you are shrinking a configuration of your own.
+
+The starting point was [everything-claude-code](https://github.com/affaan-m/ECC) (MIT); the goal was the same coverage and behaviour at the minimum tokens loaded per invocation. Everything built since — the trial harness, guards, validator, benchmark, and employee templates — is original to this project and covered in `CHANGELOG.md`.
 
 ## What Changed
 

README.md

@@ -1,25 +1,34 @@
-# Everything Claude Code — Lean
+# agent-trials
 
-[![tests](https://github.com/OsamaA140/everything-claude-code-lean/actions/workflows/tests.yml/badge.svg)](https://github.com/OsamaA140/everything-claude-code-lean/actions/workflows/tests.yml)
+[![tests](https://github.com/OsamaA140/agent-trials/actions/workflows/tests.yml/badge.svg)](https://github.com/OsamaA140/agent-trials/actions/workflows/tests.yml)
 
-Production-ready agents, skills, hooks, commands, and rules for Claude Code — a **token-optimized fork** of [affaan-m/everything-claude-code](https://github.com/affaan-m/everything-claude-code) that keeps the coverage and cuts the cost.
+**Prove an agent does its job before you trust it.** Hire-ready employee templates, a deterministic trial harness that grades agents against fixtures with planted mistakes, zero-token safety hooks, and a token-lean agent/skill set for Claude Code.
 
-## Why this fork: it barely costs tokens
+## Trials: behavioural testing for agents
 
-Most plugin stacks quietly tax every session and every invocation. This one doesn't:
+Everyone ships agents. Nobody proves they work. A **trial** hands an employee a fake business with mistakes planted in it and checks deterministically whether it caught them, stayed inside its write scope, and got the arithmetic right.
 
-Measured, not claimed — full methodology and reproduction steps in [docs/BENCHMARK.md](docs/BENCHMARK.md):
+```bash
+node scripts/trial.js prepare trials/ops-manager-smb --workspace /tmp/t1
+node scripts/trial.js report  trials/ops-manager-smb --runs /tmp/t1,/tmp/t2,/tmp/t3
+```
+
+The headline metric is **pass^k** — passed in *every* run, not merely once — because one success is capability, not reliability. No LLM judge is involved: [judge style bias runs 0.76-0.92](https://arxiv.org/abs/2604.23178), which grades prose instead of correctness. Three fixtures ship, including two held-out businesses used to separate judgement from memorisation. Methodology, measured results, and a retracted finding in [docs/TRIALS.md](docs/TRIALS.md).
+
+## Token cost: measured, not claimed
 
-| | This fork | Upstream ECC (measured @ e4e4163) |
+Most stacks quietly tax every session and every invocation. Full methodology and reproduction steps in [docs/BENCHMARK.md](docs/BENCHMARK.md):
+
+| | agent-trials | A full-size stack (measured) |
 |---|---|---|
 | Always-on context cost | **~1,260 tokens** for all 26 skills + 9 agents | ~93,000 tokens for a full checkout |
 | Avg agent invocation | **623 tokens** | 1,605 tokens (**61% less**) |
 | Avg skill invocation | **455 tokens** | 2,203 tokens (**79% less**) |
 | Avg command invocation | **189 tokens** | 970 tokens (**81% less**) |
 | Hooks (secret blocking, command guard, formatting) | **0 tokens** — they run in the harness, not the model | 0 (same mechanism) |
-| Verification | 100/100 unit tests, CI on Node 20/22, guards live-fired in real sessions | — |
+| Verification | 134 unit tests, CI on Node 20/22, guards live-fired in real sessions | — |
 
-(Fair-comparison note: upstream is a marketplace meant to be partially enabled, and covers far more ground — the per-invocation averages are the like-for-like numbers. Details and caveats in the benchmark doc.)
+(Fair-comparison note: the comparison stack is a marketplace meant to be partially enabled and covers far more ground, so the per-invocation averages are the like-for-like numbers. Details and caveats in the benchmark doc.)
 
 How: agents/skills state *when and why*, not re-teach *how* (Claude already knows the syntax); commands are thin pointers to agents instead of duplicating them; execution agents run on Sonnet with Opus reserved for architecture, planning, and security; and all enforcement lives in zero-token hooks. Full change log in `OPTIMIZATION.md` and `CHANGELOG.md`.
 
@@ -52,8 +61,8 @@ mcp-configs/ Example MCP server configs (replace YOUR_*_HERE placeholders)
 
 **As a plugin (recommended):**
 ```bash
-/plugin marketplace add <path-or-repo-you-host-this-at>
-/plugin install everything-claude-code@everything-claude-code
+/plugin marketplace add OsamaA140/agent-trials
+/plugin install agent-trials@agent-trials
 ```
 
 **Manual:**
@@ -100,5 +109,8 @@ Don't enable every MCP at once — budget ~20-30 configured, <10 enabled per pro
 node tests/run-all.js
 ```
 
-## Credit & License
-MIT, upstream by [Affaan Mustafa](https://x.com/affaanmustafa) — [original repo](https://github.com/affaan-m/everything-claude-code) (10+ months of production use, Anthropic x Forum Ventures hackathon winner). This fork only compresses/restructures for token efficiency; all credit for the underlying content and design goes upstream. Use freely, modify as needed.
+## License & Notice
+
+MIT — see [LICENSE](LICENSE). Use freely, modify as needed.
+
+The trial harness, guards library, agent validator, benchmark, employee templates, fixtures, and test suite are original to this project. Some agent, skill, command, and rule definitions, and parts of the cross-platform hook scripts, derive from [everything-claude-code](https://github.com/affaan-m/ECC) by Affaan Mustafa, used under the MIT License.

docs/BENCHMARK.md

@@ -42,7 +42,7 @@ Measured 2026-08-02. This repo v2.0.1 vs upstream [affaan-m/ECC](https://github.
 ## Reproduce it yourself
 
 ```bash
-git clone https://github.com/OsamaA140/everything-claude-code-lean && cd everything-claude-code-lean
+git clone https://github.com/OsamaA140/agent-trials && cd agent-trials
 git clone --depth 1 https://github.com/affaan-m/ECC /tmp/upstream-ecc
 node scripts/benchmark.js /tmp/upstream-ecc
 ```