feat: add model provenance tracking across plans and audits

.gitignore

@@ -15,6 +15,11 @@ venv/
 !.env.example
 
 *.log
+*.logs
+.logs
+
+# Docs
+docs/
 
 # QA / installer scratch (local-only)
 QA-Logs/
@@ -47,11 +52,11 @@ config.toml.bak
 CRUSH.md
 
 # Tests are dev-only (not needed on consumer installs)
-tests/
-!tests/
 tests/*
 !tests/test_workspace_primer_hooks.py
 !tests/test_scriptlib.py
+!tests/test_env_probe_codex.py
+!tests/test_plan_auditor_master.py
 
 # Ruler outputs at repo root (no leading dot)
 /AGENTS.md
@@ -66,7 +71,7 @@ tests/*
 !/.gitattributes
 !/.gitmodules
 !/.env.example
-# `.cursor/` stays fully ignored; ship agents/skills via config/templates/cursor-* (deploy: prime_workspace).
+# `.cursor/` stays fully ignored; ship subagents from config/templates/agents/ and skills from config/templates/cursor-skills/ (deploy: prime_workspace).
 # If your team commits other root dotdirs (e.g. .husky), add: !/.husky/
 # `.devdocs/` is covered by `/.*` above — local-only; never add ! exceptions for it.
 # END BRAINDRAIN GITIGNORE PROTOCOL

CHANGELOG.md

@@ -11,17 +11,20 @@ The format is based on keeping a clear, user-facing history. Version in `VERSION
 - **Scriptlib modularization**: scriptlib now treats project-local `.scriptlib/` and shared `~/.braindrain/scriptlib` as distinct layers, with promotion-only flow into the shared personal catalog.
 - **New scriptlib MCP tools**: added promote, update discovery/application, maintenance, and catalog status flows for local/shared script operations.
 - **Librarian-first routing**: freestanding reusable scripts are now expected to go through librarian decision flow (`reuse`, `fork`, or `new`) before a fresh script is created.
+- **Model provenance controls**: added `provenance` config toggles for chat footer scope, plan metadata stamping, and subagent model tracing (`.braindrain/plan-reports/model-trace.jsonl`), plus audit report frontmatter fields for model/date/cursor mode attribution.
+- **Cursor stop hook stability**: `on-stop-observe.sh` is now output-silent by default so Cursor stop-hook JSON parsing is not broken by plain-text stdout.
 
 ### For contributors
 
-- **Cursor agents/skills**: librarian and `scriptlib-librarian` live under `config/templates/cursor-subagents/` and `config/templates/cursor-skills/`; `prime_workspace()` deploys both to `.cursor/agents/` and `.cursor/skills/`. The repo root gitignore no longer whitelists paths under `.cursor/` (Braindrain protocol: ship via templates only).
+- **Subagent templates**: single source tree `config/templates/agents/` deploys to `.cursor/agents/` and/or `.codex/agents/` depending on the resolved IDE set; duplicate `cursor-subagents/` and `codex-subagents/` template dirs were removed. Skills remain under `config/templates/cursor-skills/`. Added `daily-plan-auditor` agent and planning close-out guidance in Ruler `RULES.md`. Planning audit script moves `archived` plans into `<ide>/plans/.plan.archives/`. Tests: `tests/test_plan_auditor_master.py`.
+- Added provenance-aware runtime/tooling paths in `braindrain/server.py`, `braindrain/config.py`, `braindrain/types.py`, and `scripts/daily_plan_audit.py`, with tests in `tests/test_plan_auditor_master.py` and `tests/test_provenance_config.py`.
 
 ## [1.0.3] — 2026-04-10
 
 ### For users
 
 - You can opt into **scriptlib**: harvest and search reusable scripts, run them through MCP, and keep guidance in agent rules when scriptlib is enabled for a workspace.
-- **prime_workspace** now deploys **Cursor and Codex subagent templates** from `config/templates/cursor-subagents/` and `config/templates/codex-subagents/`, with optional `sync_subagents` and `codex_agent_targets` for Codex layout.
+- **prime_workspace** deploys Cursor/Codex subagent markdown into `.cursor/agents/` and `.codex/agents/` with optional `sync_subagents` and `codex_agent_targets`. (Older docs referred to split template trees; the repo now uses a single canonical tree — see Unreleased.)
 - **Token checkpoint protocol** and optional `.braindrain/token-metrics.jsonl` schema `1.0` are documented in templates and README for consistent observability.
 
 ### For contributors

TRADEMARKS.md

@@ -4,7 +4,7 @@ This policy applies to the project name **Braindrain**, any official project log
 
 ### Code license vs. brand usage
 
-The source code is licensed under `AGPL-3.0` in [`LICENSE`](LICENSE).
+The source code is licensed under `AGPL-3.0` in `[LICENSE](LICENSE)`.
 That license governs code use, modification, and distribution.
 This policy is only about avoiding confusion around the official project identity.
 
@@ -29,4 +29,4 @@ If you ship a fork or service, please use distinct branding unless you have expl
 
 ### Permission requests
 
-If you want to use the Braindrain name or branding in a way that may look official, open an issue in this repository with the intended use and context.
+If you want to use the Braindrain name or branding in a way that may look official, open an issue in this repository with the intended use and context.

braindrain/config.py

@@ -107,6 +107,7 @@ def _parse_config(self, raw: dict) -> ConfigData:
             lessons=raw.get("lessons", {}),
             dreaming=raw.get("dreaming", {}),
             provider_context=raw.get("provider_context", {}),
+            provenance=raw.get("provenance", {}),
         )
 
     def reload(self) -> None:

braindrain/env_probe.py

@@ -343,9 +343,10 @@
 
 # ---------------------------------------------------------------------------
 # IDE / Agent MCP config locations
-# Checked by path existence; content parsed if JSON/JSONC.
+# Checked by path existence; content parsed if JSON/JSONC, except Codex CLI
+# which is TOML-first at ~/.codex/config.toml.
 # Format: (app_key, display_name, config_path, mcp_key_path)
-#   mcp_key_path: dot-separated path into the JSON to find server map
+#   mcp_key_path: dot-separated path into the config to find server map
 #   e.g. "mcpServers" or "mcp" or "context_servers"
 # ---------------------------------------------------------------------------
 
@@ -374,7 +375,7 @@
         "~/.config/claude/claude_desktop_config.json",
         "mcpServers",
     ),
-    ("codex_cli", "Codex CLI", "~/.codex/config.json", "mcpServers"),
+    ("codex_cli", "Codex CLI", "~/.codex/config.toml", "mcp_servers"),
     ("codex_openai", "Codex (OpenAI)", "~/.openai/mcp.json", "mcpServers"),
     ("continue", "Continue", "~/.continue/config.json", "mcpServers"),
     ("vscode", "VS Code", "~/.vscode/settings.json", "mcp.servers"),
@@ -423,6 +424,18 @@ def _read_json_file(path: Path) -> dict | None:
         return None
 
 
+def _read_toml_file(path: Path) -> dict | None:
+    """Read a TOML file, return parsed dict or None."""
+    try:
+        import tomllib  # Python 3.11+ stdlib
+
+        raw = path.read_text(encoding="utf-8", errors="ignore")
+        parsed = tomllib.loads(raw)
+        return parsed if isinstance(parsed, dict) else None
+    except Exception:
+        return None
+
+
 def _get_nested(d: dict, dot_path: str) -> Any:
     """Walk a dot-separated key path into a nested dict."""
     parts = dot_path.split(".")
@@ -456,7 +469,7 @@ def probe_app_configs() -> dict[str, Any]:
             }
             continue
 
-        parsed = _read_json_file(path)
+        parsed = _read_toml_file(path) if app_key == "codex_cli" else _read_json_file(path)
         mcp_block = _get_nested(parsed, mcp_key) if parsed else None
 
         server_names: list[str] = []