Merge branch 'main' into reflective

# Conflicts:
#	README.md
#	orchestrator/iteration.py
#	orchestrator/schemas/campaign.schema.yaml

Author: sriumcp

.claude/commands/index-wiki.md

@@ -0,0 +1,107 @@
+Merge a single campaign's extracted knowledge into the cross-campaign registry.
+
+## Usage
+
+`/index-wiki <campaign-name>` — indexes one campaign into registry.json
+
+## Steps
+
+1. **Resolve campaign**: Use `$ARGUMENTS` as the campaign name. Verify `~/.nous/wiki/campaigns/<campaign-name>/concepts.json` exists.
+
+2. **Read per-campaign files**: From `~/.nous/wiki/campaigns/<campaign-name>/`:
+   - `concepts.json` — contains `repo_path`, `system_name`, `research_question`, `campaign_name`, `date`, plus `entities`, `concepts`, `parameters`
+   - `principles.json` — full principle definitions (extract IDs for the registry)
+   - `dead-ends.json` — refuted approaches
+   - `frontiers.json` — boundary conditions
+   - `interactions.json` — untested combinations
+
+3. **Load existing registry**: Read `~/.nous/wiki/registry.json`. If it doesn't exist, initialize:
+   ```json
+   {"version": 1, "projects": {}}
+   ```
+
+4. **Idempotency check**: Look for this campaign in `projects[repo_path].campaigns[]` by name. If already present, report "Campaign already in registry — skipping" and stop.
+
+5. **Find or create project**: Look up `projects[repo_path]`. If not present, create:
+   ```json
+   {
+     "name": "<system_name from concepts.json>",
+     "campaigns": [],
+     "entities": []
+   }
+   ```
+
+6. **Process entities**: For each entity in `concepts.json`:
+   - Normalize the name (lowercase, strip content in parentheses) for matching
+   - Check if an entity with the same normalized name exists in `projects[repo_path].entities[]`
+     - If yes: add this campaign name to its `campaigns[]` array (if not already there)
+     - If no: assign new ID `E-{max_entity_id + 1}`, create entry:
+       ```json
+       {"id": "E-N", "name": "<entity name>", "aliases": [], "campaigns": ["<campaign-name>"]}
+       ```
+   - Track the max entity ID across all projects globally (IDs are globally unique)
+
+7. **Build campaign object**: Create the campaign entry for the registry:
+   ```json
+   {
+     "name": "<campaign_name>",
+     "date": "<date>",
+     "research_question": "<research_question>",
+     "concepts": [],
+     "parameters": [],
+     "principles": [],
+     "dead_ends": [],
+     "frontiers": [],
+     "interactions": []
+   }
+   ```
+
+   - **Concepts**: For each concept in `concepts.json`, assign ID `C-{max+1}` and add `{"id": "C-N", "name": "<concept name>"}` to the campaign's `concepts[]`
+   - **Parameters**: For each parameter in `concepts.json`, assign ID `P-{max+1}` and add `{"id": "P-N", "name": "<parameter name>"}` to the campaign's `parameters[]`
+   - **Principles**: Read `principles.json`, add each principle's ID string (e.g., "RP-1") to the campaign's `principles[]` array
+   - **Dead-ends**: For each entry in `dead-ends.json`, assign ID `DE-{max+1}` and add `{"id": "DE-N", "title": "<title>"}` to the campaign's `dead_ends[]`
+   - **Frontiers**: For each entry in `frontiers.json`, assign ID `F-{max+1}` and add `{"id": "F-N", "title": "<title>"}` to the campaign's `frontiers[]`
+   - **Interactions**: For each entry in `interactions.json`, assign ID `I-{max+1}` and add `{"id": "I-N", "title": "<title>"}` to the campaign's `interactions[]`
+
+8. **Add campaign to project**: Append the campaign object to `projects[repo_path].campaigns[]`.
+
+9. **Recompute entity clusters**: Using ALL entities in `projects[repo_path].entities[]` and their definitions from per-campaign `concepts.json` files:
+
+   a. For each entity, gather its definition from the campaign(s) that define it (read `~/.nous/wiki/campaigns/<campaign>/concepts.json` → `entities[]` → match by name). If an entity appears in multiple campaigns, concatenate definitions with ` | `.
+
+   b. Semantically group entities by functional role/purpose into clusters. Rules:
+      - Min 2 entities per cluster, max 10
+      - Max 20 clusters total per project — if grouping would exceed 20, merge the smallest/most-similar clusters
+      - Each entity belongs to at most one cluster (singletons are valid — omit them)
+      - Labels: 2-4 words, Title Case, describe the functional group
+
+   c. Assign sequential IDs `EC-1`, `EC-2`, etc. Set `projects[repo_path].entity_clusters` to:
+      ```json
+      [{"id": "EC-1", "label": "...", "entities": ["E-4", "E-5", "E-6"]}, ...]
+      ```
+
+   d. **CRITICAL**: This step ONLY writes the `entity_clusters` field. Do NOT modify `name`, `campaigns`, `entities`, or any other field on the project or registry.
+
+10. **Write registry.json**: Write the updated registry to `~/.nous/wiki/registry.json` with 2-space indentation.
+
+11. **Report**: Print the registry path and a summary of what was added (counts of entities, concepts, parameters, principles, dead-ends, frontiers, interactions).
+
+## ID Assignment Rules
+
+- IDs are **globally unique** across all projects and campaigns
+- To find the next ID for a type (e.g., "C-"), scan ALL projects and campaigns in the registry for the maximum existing ID of that type, then increment
+- Entity IDs (E-N) are unique per-project in practice but globally unique in assignment
+- Principle IDs (RP-N) are NOT reassigned — they keep their original campaign-local IDs
+
+## Deduplication Rules
+
+- **Entities**: Match by normalized name (lowercase, strip parenthetical suffixes). If matched, add campaign to existing entity's `campaigns[]` array.
+- **Everything else** (concepts, parameters, dead-ends, frontiers, interactions): Always create new entries. Different campaigns may have similar-sounding entries that are contextually distinct. Never deduplicate these across campaigns.
+
+## Important Rules
+
+- This skill only reads from `~/.nous/wiki/campaigns/` — it never reaches back to source repos
+- This is the ONLY skill that writes to `registry.json`
+- All metadata (`repo_path`, `system_name`, `research_question`) comes from the per-campaign `concepts.json`, which was populated by `/post-campaign` from the source `campaign.yaml`
+- Process one campaign at a time (read registry → modify → write) to prevent conflicts
+- `entity_clusters` is the ONLY registry field that is recomputed (not appended). It is fully replaced each time a new campaign is indexed. All other data is strictly append-only.