feat(F097): Product Truth and Activation Loop

.claude/commands.json

@@ -92,6 +92,14 @@
       "description": "Codebase analysis",
       "phase": "analysis"
     },
+    "strataudit": {
+      "file": "skills/strataudit.md",
+      "cli": "sdp-strataudit run",
+      "mode": "hybrid",
+      "llm_subagents": ["analyst", "architect"],
+      "description": "Evidence-backed strategy traceability audit",
+      "phase": "analysis"
+    },
     "feature": {
       "file": "skills/feature.md",
       "cli": "sdp plan",

.gitignore

@@ -62,8 +62,15 @@ archive/
 !.sdp/evidence/
 !.sdp/checkpoints/
 .sdp/log/events.jsonl
+.sdp/log/events.jsonl.lock
+.sdp/state.json
+.sdp/review_verdict.json
+.sdp/tier_metrics.json
 sdp-plugin/.sdp/log/events.jsonl
 **/events.jsonl
+**/events.jsonl.lock
+.beads/.jsonl.lock
+sdp-plugin/.beads/.jsonl.lock
 
 # Build output
 bin/

.opencode/commands/strataudit.md

@@ -0,0 +1,35 @@
+---
+description: Evidence-backed strategy traceability audit over a document corpus; use when the user needs document-grounded alignment analysis across strategy, architecture, design, or implementation materials. Prefer an injected host-native runtime when available, otherwise use a configured OpenAI-compatible runtime; OpenRouter is the default network accelerator, not the only path.
+agent: architect
+---
+
+# /strataudit — StratAudit
+
+## Overview
+
+This command implements the `strataudit` skill from the SDP workflow.
+
+See `/prompts/skills/strataudit/SKILL.md` for complete documentation.
+
+## Usage
+
+```bash
+/strataudit [arguments]
+```
+
+## Implementation
+
+The command delegates to the `strataudit` skill, which provides:
+
+- mode-based document-backed audit flow
+- explicit runtime selection order
+- structured artifact output
+- trust-oriented failure and refusal behavior
+
+## Related
+
+- Skills: `prompts/skills/strataudit/SKILL.md`
+- Reference: `docs/reference/strataudit-evidence-policy.md`
+- Reference: `docs/reference/strataudit-runtime-policy.md`
+- Reference: `docs/reference/strataudit-output-modes.md`
+- Agents: `prompts/agents/architect.md`

CLAUDE.md

@@ -10,7 +10,7 @@ Quick reference for using SDP CLI v0.9.8 with Claude Code.
 @vision "AI-powered task manager"    # Strategic planning
 @reality --quick                     # Codebase analysis
 @feature "Add user authentication"   # Plan feature
-@build 00-001-01                     # Execute workstream
+@build 00-001-01                     # Execute executable leaf workstream
 @review <feature-id>                 # Quality check
 ```
 
@@ -83,9 +83,9 @@ New project?
 
 **@ux** — UX research for user-facing features (standalone or auto-triggered by @feature)
 
-**@oneshot** — Workstreams exist, want autonomous execution with checkpoint/resume
+**@oneshot** — Workstream tree exists, want autonomous execution of ready leaf workstreams with checkpoint/resume
 
-**@build** — Execute a single workstream (use instead of @oneshot for 1-2 WS)
+**@build** — Execute a single executable leaf workstream (use instead of @oneshot for 1-2 leaf WS)
 
 ---
 
@@ -103,7 +103,7 @@ New project?
 | `@ux` | UX research (mental model elicitation) | Planning |
 | `@design` | Workstream design (EnterPlanMode) | Planning |
 | `@oneshot` | Execution orchestrator (autonomous) | Execution |
-| `@build` | Execute single workstream (TDD) | Execution |
+| `@build` | Execute single executable leaf workstream (TDD) | Execution |
 | `@review` | Multi-agent quality review | Execution |
 | `@deploy` | Merge feature branch to main | Execution |
 
@@ -166,14 +166,14 @@ New project?
 # 1. Plan feature
 @feature "Add payment processing"
 
-# 2. Execute all workstreams
+# 2. Execute all ready leaf workstreams
 @oneshot <feature-id>
 ```
 
 ### Manual Flow (learning or debugging)
 
 ```bash
-@build 00-050-01   # Execute one at a time
+@build 00-050-01   # Execute one leaf at a time
 @build 00-050-02
 @review <feature-id>       # Review when done
 @deploy <feature-id>       # Deploy
@@ -188,8 +188,9 @@ New project?
    - [PROTOCOL.md](docs/PROTOCOL.md)
 
 2. **Key concepts:**
-   - **Workstream (WS)**: Atomic task, one-shot execution
-   - **Feature**: 5-30 workstreams
+   - **Aggregate Workstream**: non-executable container or roll-up over 2+ leaf workstreams
+   - **Leaf Workstream**: atomic executable unit
+   - **Feature**: 5-30 workstreams total, but only leaves are directly executable
    - **Release**: 10-30 features
 
 3. **Install Beads CLI** (task tracking):

README.md

@@ -83,6 +83,7 @@ Canonical prompt sources live in `prompts/`. Tool-specific directories such as `
 | [QUICKSTART.md](docs/QUICKSTART.md) | Recommended first-success path |
 | [CLI_REFERENCE.md](docs/CLI_REFERENCE.md) | Current `sdp` command surfaces |
 | [PROTOCOL.md](docs/PROTOCOL.md) | Current protocol overview |
+| [PRODUCT_CONTRACT.md](docs/PRODUCT_CONTRACT.md) | Product definition and mode policy |
 | [reference/README.md](docs/reference/README.md) | Reference index and legacy-doc status |
 | [.codex/INSTALL.md](.codex/INSTALL.md) | Codex-specific install notes |
 | [MANIFESTO.md](docs/MANIFESTO.md) | Vision and rationale |