Larens94 · cursor · Sep 8, 2026 · Sep 8, 2026 · Sep 8, 2026 · Sep 8, 2026
diff --git a/.codedna b/.codedna
@@ -30,14 +30,6 @@ packages:
 cross_cutting_patterns: {}
 
 agent_sessions:
-  - agent: "gpt-5"
-    provider: "openai"
-    date: "2026-08-21"
-    session_id: "s_20260821_gitbook_docs"
-    task: "Redesign documentation like GitBook"
-    changed: ["docs/install.html","docs/site.css",".codedna"]
-    visited: ["docs/install.html","docs/site.css",".codedna"]
-    message: "Reworked the documentation shell with a GitBook-inspired three-column layout, searchable hierarchical sidebar, keyboard shortcut, breadcrumb, page metadata, sticky outline, pagination, feedback, and responsive spacing. HTML/CSS checks, Ruff, and 284 tests pass."
   - agent: "gpt-5"
     provider: "openai"
     date: "2026-08-21"
@@ -70,3 +62,11 @@ agent_sessions:
     changed: ["codedna_tool/languages/php.py","codedna_tool/languages/base.py","codedna_tool/languages/vbnet.py","codedna_tool/languages/__init__.py","codedna_tool/cli.py","integrations/.roorules","integrations/install.sh","integrations/README.md","docs/languages.md","README.md","README-it.md","CHANGELOG.md","tests/test_issues_4_6_7.py","tests/test_docs.py"]
     visited: ["codedna_tool/languages/php.py","codedna_tool/languages/base.py","codedna_tool/languages/csharp.py","codedna_tool/languages/__init__.py","codedna_tool/cli.py","integrations/.clinerules"]
     message: "Validated #4/#6/#7 still open. Fixed PHP mixed-HTML injection + full-file has_codedna_header; added VbNetAdapter; added Roo --tools roo. 306 tests pass."
+  - agent: "composer-2.5"
+    provider: "cursor"
+    date: "2026-09-08"
+    session_id: "s_20260908_001"
+    task: "Challenge honesty + experimental notes"
+    changed: ["docs/challenge.md","docs/challenge.it.md","docs/challenge.html","challenge/SUBMISSION_TEMPLATE.md","challenge/SUBMISSION_TEMPLATE.it.md","README.md","README-it.md"]
+    visited: ["docs/challenge.md","docs/challenge.it.md","docs/challenge.html","challenge/SUBMISSION_TEMPLATE.md","challenge/SUBMISSION_TEMPLATE.it.md","README.md","README-it.md"]
+    message: "Documented experimental agent×language caveat + honesty/Meet review + public lives in challenge rules."
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -6,7 +6,7 @@ All notable changes to CodeDNA will be documented in this file.
 
 ### Added
 
-- **CodeDNA Challenge (€200)** — bilingual rules, required `metrics.json` (languages + frameworks + size/files), docs site page + static board (`docs/challenge.html`, `docs/challenge-board.json`). PR opens enrollment (no separate signup). Separate from historical SWE-bench benchmarks.
+- **CodeDNA Challenge (€200)** — bilingual rules, required `metrics.json` (languages + frameworks + size/files ≥25 + `install` + `setup.layout` + `bugs_reported`), optional file-localization metrics + optional judge-agent comparison, real-project bar (no toy sites), same-task control vs CodeDNA via two branches/checkouts/twin projects, copy-paste agent/judge prompts + task list, Meet presentation verification, docs site page + live public board (IT default) updated as valid PRs arrive. PR opens enrollment (no separate signup). Separate from historical SWE-bench benchmarks.
 
 ### Added
 

diff --git a/README-it.md b/README-it.md
@@ -48,9 +48,9 @@ Nessuna infrastruttura. Nessuna pipeline di retrieval. Nessuna memoria esterna.
 
 ## CodeDNA Challenge — €200
 
-Valutazione pubblica sul **tuo** progetto (non SWE-bench). Almeno 10 task con/senza CodeDNA, stack L0/L1/L2 a pari livello, PR con metriche — anche se vanno contro CodeDNA.
+Valutazione pubblica sul **tuo** progetto (non SWE-bench). Progetto reale funzionante (≥25 file), stack tecnologico + install CodeDNA dichiarati, stessa metodologia per tutti (due branch / checkout / progetti gemelli), ≥10 **stessi** task con/senza CodeDNA, [prompt da copiare per l’agente](docs/challenge-agent-prompt.it.md), PR con metriche — anche se vanno contro CodeDNA. I bug vanno segnalati (sperimentale tra agenti). Possibile presentazione Meet; la bacheca pubblica si aggiorna quando arrivano PR valide.
 
-- Bacheca: [larens94.github.io/codedna/challenge.html](https://larens94.github.io/codedna/challenge.html)
+- Classifica / bacheca: [larens94.github.io/codedna/challenge.html](https://larens94.github.io/codedna/challenge.html)
 - Regolamento: [`docs/challenge.it.md`](docs/challenge.it.md) (IT) · [`docs/challenge.md`](docs/challenge.md) (EN)
 - Submission: `challenge/<handle>/metrics.json` ([esempio](challenge/metrics.example.json)) — **aprire la PR ti iscrive**
 

diff --git a/README.md b/README.md
@@ -48,9 +48,9 @@ No infrastructure. No retrieval pipeline. No external memory. The code carries i
 
 ## CodeDNA Challenge — €200
 
-Public **own-project** evaluation (not SWE-bench). ≥10 tasks with/without CodeDNA, fair L0/L1/L2 stack parity, metrics PR — even if results go against CodeDNA.
+Public **own-project** evaluation (not SWE-bench). Real working project (≥25 files), declared tech stack + CodeDNA install, same methodology for everyone (two branches / checkouts / twin projects), ≥10 **same** tasks with/without CodeDNA, copy-paste [agent prompt](docs/challenge-agent-prompt.md), metrics PR — even if results go against CodeDNA. Bugs must be reported (experimental across agents). Meet presentation may be required; public board updates as valid PRs arrive.
 
-- Board: [larens94.github.io/codedna/challenge.html](https://larens94.github.io/codedna/challenge.html)
+- Ranking / board: [larens94.github.io/codedna/challenge.html](https://larens94.github.io/codedna/challenge.html)
 - Rules: [`docs/challenge.md`](docs/challenge.md) (EN) · [`docs/challenge.it.md`](docs/challenge.it.md) (IT)
 - Submission: `challenge/<handle>/metrics.json` ([example](challenge/metrics.example.json)) — **opening the PR enrolls you**
 

diff --git a/challenge/AGENT_PROMPT.md b/challenge/AGENT_PROMPT.md
@@ -0,0 +1,13 @@
+# Agent prompts (copy-paste)
+
+Canonical prompts (also served on GitHub Pages):
+
+- Italiano: [`docs/challenge-agent-prompt.it.md`](../docs/challenge-agent-prompt.it.md)
+- English: [`docs/challenge-agent-prompt.md`](../docs/challenge-agent-prompt.md)
+
+Task checklist:
+
+- Italiano: [`TASKS_TEMPLATE.it.md`](./TASKS_TEMPLATE.it.md)
+- English: [`TASKS_TEMPLATE.md`](./TASKS_TEMPLATE.md)
+
+On the public page use **Copia per il tuo agente** → https://larens94.github.io/codedna/challenge.html
diff --git a/challenge/README.md b/challenge/README.md
@@ -4,19 +4,30 @@
 
 ```text
 challenge/<github-handle>/
-  metrics.json    # REQUIRED
+  metrics.json    # REQUIRED (stack + install + setup.layout + same-task control/codedna results)
   README.md       # optional
   notes.md        # optional
 ```
 
-No official signup. Public board (maintainer-updated): [`docs/challenge.html`](../docs/challenge.html) · data [`docs/challenge-board.json`](../docs/challenge-board.json)
+No official signup. Public ranking (updated as valid PRs arrive): [`docs/challenge.html`](../docs/challenge.html) · data [`docs/challenge-board.json`](../docs/challenge-board.json)
+
+**Copy for your agent**
+
+- Italiano: [`docs/challenge-agent-prompt.it.md`](../docs/challenge-agent-prompt.it.md)
+- English: [`docs/challenge-agent-prompt.md`](../docs/challenge-agent-prompt.md)
+- Judge (optional, after both sessions): [`docs/challenge-judge-prompt.it.md`](../docs/challenge-judge-prompt.it.md) · [EN](../docs/challenge-judge-prompt.md)
+- Task lists: [`TASKS_TEMPLATE.it.md`](./TASKS_TEMPLATE.it.md) · [`TASKS_TEMPLATE.md`](./TASKS_TEMPLATE.md)
+- Pointer: [`AGENT_PROMPT.md`](./AGENT_PROMPT.md)
+
+Methodology (same for everyone): same ≥10 tasks with vs without CodeDNA via `two_branches` / `two_checkouts` / `two_projects`.  
+File-localization fields (`files_expected`, F1, …) are **optional** — use when known; otherwise a judge agent can still compare sessions.
 
 **Metrics format**
 
 - Example: [`metrics.example.json`](./metrics.example.json)
 - Schema: [`metrics.schema.json`](./metrics.schema.json)
 
-Required project fields: `languages`, `frameworks`, `approx_source_files`, `size_band` (`S`/`M`/`L`/`XL`).
+Required: `languages`, `frameworks`, `approx_source_files` (≥25), `size_band`, `install`, `setup.layout`, `bugs_reported`.
 
 **Rules**
 

diff --git a/challenge/SUBMISSION_TEMPLATE.it.md b/challenge/SUBMISSION_TEMPLATE.it.md
@@ -7,36 +7,59 @@
 ### Partecipante
 
 - Handle:
-- Issue di iscrizione collegata:
 - Modalità challenge:
   - [ ] Parity (stessi L1/L2 su Control e CodeDNA)
   - [ ] Dichiarata `codedna-only` vs stack superiore
 
-### Stack sotto test
+### Stack tecnologico (obbligatorio)
+
+- Linguaggio/i:
+- Framework (obbligatorio — non solo il linguaggio):
+- Agente / modello:
+- File sorgente approx. (≥ **25**):
+- Size band (`S`/`M`/`L`/`XL`):
+- Note stack (DB, monorepo, infra — opzionale):
+- URL pubblico (opzionale) / “privato — solo metriche”:
+- [ ] È un **progetto reale e funzionante** (non sito giocattolo / hello-world / demo throwaway)
+
+### Install CodeDNA (obbligatorio)
+
+- Agente usato con CodeDNA:
+- Passi esatti di install / init (incolla i comandi):
+- Valore di `codedna install --tools` (se usato):
+- Install + annotation ok? sì / no — note:
+
+### Setup Control vs CodeDNA (obbligatorio)
+
+- [ ] `two_branches` (es. `challenge/control` + `challenge/codedna`)
+- [ ] `two_checkouts` (due cartelle)
+- [ ] `two_projects` (progetti gemelli)
+- Ref Control / CodeDNA:
+- [ ] Prompt agente usato: [`docs/challenge-agent-prompt.it.md`](../docs/challenge-agent-prompt.it.md)
+- [ ] Lista task congelata: [`TASKS_TEMPLATE.it.md`](./TASKS_TEMPLATE.it.md)
+
+### Stack sotto test (L0/L1/L2)
 
 | Livello | Tool / file | Presente in Control? | Presente in CodeDNA? |
 |---|---|---|---|
 | L0 CodeDNA | header in-source | no | sì |
 | L1 | | | |
 | L2 | | | |
 
-### Progetto
-
-- Linguaggio/i:
-- Framework (obbligatorio — non solo il linguaggio):
-- File sorgente approx.:
-- Size band (`S`/`M`/`L`/`XL`):
-- URL pubblico (opzionale) / “privato — solo metriche”:
-
 ### Metrics JSON (obbligatorio)
 
 - [ ] Presente `challenge/<handle>/metrics.json`
 - [ ] Partito da [`metrics.example.json`](./metrics.example.json) / conforme a [`metrics.schema.json`](./metrics.schema.json)
 - [ ] `schema_version` = `"1.0"`
-- [ ] Compilati `project.languages` + `project.frameworks` + `approx_source_files` + `size_band`
+- [ ] Compilati `project.languages` + `project.frameworks` + `approx_source_files` (≥25) + `size_band`
+- [ ] Compilati `install.agent` + `install.steps`
+- [ ] Compilato `setup.layout` (`two_branches` | `two_checkouts` | `two_projects`)
+- [ ] Presente `bugs_reported` (array vuoto OK se nessuno)
 - [ ] ≥10 task con mix `difficulty` (easy ≥3, medium ≥3, hard ≥2)
-- [ ] Ogni task ha risultati `control` e `codedna`
+- [ ] Gli **stessi task** hanno risultati `control` e `codedna`
 - [ ] `summary.favors` impostato (`codedna` | `control` | `tie` | `inconclusive`)
+- [ ] (Opzionale) `files_expected` / file F1 solo se conosciuti — non inventati
+- [ ] (Opzionale) giudice usato: [`docs/challenge-judge-prompt.it.md`](../docs/challenge-judge-prompt.it.md) → `judge` / `tasks[].judge`
 - [ ] I risultati possono favorire **oppure** sfavorire CodeDNA (onestà OK)
 
 ### Opzionale
@@ -47,9 +70,13 @@
 ### Bug trovati in CodeDNA
 
 - Link (oppure “nessuno”) — elencarli anche in `metrics.json` → `bugs_reported`:
+- [ ] Se qualcosa si è rotto su agente/linguaggio, ho aperto issue o PR di fix
 
 ### Checklist
 
 - [ ] Nessun secret / sorgente proprietaria in questa PR
 - [ ] Solo log redatti
 - [ ] Modalità + parity di stack dichiarate nel JSON (`mode`, `stack`)
+- [ ] Metriche e narrativa sono veritiere (non inventate)
+- [ ] Posso presentare progetto e processo di test in una call di review (Meet / simile) se richiesto
+- [ ] Capisco che claim inventati o progetti finti = squalifica
diff --git a/challenge/SUBMISSION_TEMPLATE.md b/challenge/SUBMISSION_TEMPLATE.md
@@ -7,36 +7,59 @@
 ### Entrant
 
 - Handle:
-- Linked entry issue:
 - Challenge mode:
   - [ ] Parity (same L1/L2 on Control and CodeDNA)
   - [ ] Declared `codedna-only` vs higher stack
 
-### Stack under test
+### Tech stack (required)
+
+- Language(s):
+- Framework(s) (required — not language alone):
+- Agent / model:
+- Approx. source files (≥ **25**):
+- Size band (`S`/`M`/`L`/`XL`):
+- Tech stack notes (DB, monorepo, infra — optional):
+- Public URL (optional) / “private — metrics only”:
+- [ ] This is a **real working project** (not a toy site / hello-world / throwaway demo)
+
+### CodeDNA install (required)
+
+- Agent used with CodeDNA:
+- Exact install / init steps (paste commands):
+- `codedna install --tools` value (if any):
+- Did install + annotation work? yes / no — notes:
+
+### Control vs CodeDNA setup (required)
+
+- [ ] `two_branches` (e.g. `challenge/control` + `challenge/codedna`)
+- [ ] `two_checkouts` (two folders)
+- [ ] `two_projects` (twin projects)
+- Control / CodeDNA refs:
+- [ ] Agent prompt used: [`docs/challenge-agent-prompt.md`](../docs/challenge-agent-prompt.md)
+- [ ] Frozen task list: [`TASKS_TEMPLATE.md`](./TASKS_TEMPLATE.md)
+
+### Stack under test (L0/L1/L2)
 
 | Level | Tools / files | Present in Control? | Present in CodeDNA? |
 |---|---|---|---|
 | L0 CodeDNA | in-source headers | no | yes |
 | L1 | | | |
 | L2 | | | |
 
-### Project
-
-- Language(s):
-- Framework(s) (required — not language alone):
-- Approx. source files:
-- Size band (`S`/`M`/`L`/`XL`):
-- Public URL (optional) / “private — metrics only”:
-
 ### Metrics JSON (required)
 
 - [ ] `challenge/<handle>/metrics.json` present
 - [ ] Copied from [`metrics.example.json`](./metrics.example.json) / matches [`metrics.schema.json`](./metrics.schema.json)
 - [ ] `schema_version` = `"1.0"`
-- [ ] `project.languages` + `project.frameworks` + `approx_source_files` + `size_band` filled
+- [ ] `project.languages` + `project.frameworks` + `approx_source_files` (≥25) + `size_band` filled
+- [ ] `install.agent` + `install.steps` filled
+- [ ] `setup.layout` filled (`two_branches` | `two_checkouts` | `two_projects`)
+- [ ] `bugs_reported` present (empty array OK if none)
 - [ ] ≥10 tasks with `difficulty` mix (easy ≥3, medium ≥3, hard ≥2)
-- [ ] Each task has both `control` and `codedna` results
+- [ ] **Same tasks** each have both `control` and `codedna` results
 - [ ] `summary.favors` set (`codedna` | `control` | `tie` | `inconclusive`)
+- [ ] (Optional) `files_expected` / file F1 only when known — never invented
+- [ ] (Optional) judge used: [`docs/challenge-judge-prompt.md`](../docs/challenge-judge-prompt.md) → `judge` / `tasks[].judge`
 - [ ] Results may favor **or** disfavor CodeDNA (honesty OK)
 
 ### Optional
@@ -47,9 +70,13 @@
 ### Bugs found in CodeDNA
 
 - Links (or “none”) — also list them in `metrics.json` → `bugs_reported`:
+- [ ] If something broke on my agent/language, I filed an issue or fix PR
 
 ### Checklist
 
 - [ ] No secrets / proprietary source in this PR
 - [ ] Redacted logs only
 - [ ] Mode + stack parity declared in JSON (`mode`, `stack`)
+- [ ] Metrics and narrative are truthful (not invented)
+- [ ] I can present the project and test process on a review call (Meet / similar) if asked
+- [ ] I understand fabricated claims or fake projects = disqualification
diff --git a/challenge/TASKS_TEMPLATE.it.md b/challenge/TASKS_TEMPLATE.it.md
@@ -0,0 +1,35 @@
+# CodeDNA Challenge — template lista task
+
+> Congela questa lista **prima** delle run cronometrate. Esegui gli **stessi** ID su Control e su CodeDNA.  
+> Prompt completo per l’agente: [`docs/challenge-agent-prompt.it.md`](../docs/challenge-agent-prompt.it.md)
+
+## Setup (scegline uno)
+
+- [ ] `two_branches` — es. `challenge/control` + `challenge/codedna`
+- [ ] `two_checkouts` — due cartelle dello stesso progetto
+- [ ] `two_projects` — progetti gemelli dallo stesso codice di partenza
+
+Agente / modello: _______________  
+Note layout: _______________
+
+## Task (≥10 — easy ≥3, medium ≥3, hard ≥2)
+
+| ID | Difficoltà | Titolo (specifico del progetto) | Control ok? | CodeDNA ok? | Note |
+|---|---|---|---|---|---|
+| E1 | easy | | | | |
+| E2 | easy | | | | |
+| E3 | easy | | | | |
+| M1 | medium | | | | |
+| M2 | medium | | | | |
+| M3 | medium | | | | |
+| H1 | hard | | | | |
+| H2 | hard | | | | |
+| T9 | easy/medium/hard | | | | |
+| T10 | easy/medium/hard | | | | |
+
+Extra opzionali: E4, M4, H3…
+
+## Dopo le run
+
+- [ ] `metrics.json` compilato (`install`, `setup.layout`, `bugs_reported`, `summary.favors`)
+- [ ] PR aperta con submission template
diff --git a/challenge/TASKS_TEMPLATE.md b/challenge/TASKS_TEMPLATE.md
@@ -0,0 +1,35 @@
+# CodeDNA Challenge — task list template
+
+> Freeze this list **before** timed runs. Run the **same** IDs on Control and on CodeDNA.  
+> Italiano: adatta i titoli al tuo repo; scheletro anche in [`docs/challenge-agent-prompt.it.md`](../docs/challenge-agent-prompt.it.md).
+
+## Setup (pick one)
+
+- [ ] `two_branches` — e.g. `challenge/control` + `challenge/codedna`
+- [ ] `two_checkouts` — two folders of the same project
+- [ ] `two_projects` — twin projects from the same starting code
+
+Agent / model: _______________  
+Layout notes: _______________
+
+## Tasks (≥10 — easy ≥3, medium ≥3, hard ≥2)
+
+| ID | Difficulty | Title (project-specific) | Control pass? | CodeDNA pass? | Notes |
+|---|---|---|---|---|---|
+| E1 | easy | | | | |
+| E2 | easy | | | | |
+| E3 | easy | | | | |
+| M1 | medium | | | | |
+| M2 | medium | | | | |
+| M3 | medium | | | | |
+| H1 | hard | | | | |
+| H2 | hard | | | | |
+| T9 | easy/medium/hard | | | | |
+| T10 | easy/medium/hard | | | | |
+
+Optional extras: E4, M4, H3…
+
+## After runs
+
+- [ ] `metrics.json` filled (`install`, `setup.layout`, `bugs_reported`, `summary.favors`)
+- [ ] PR opened with submission template