@lifeaitools/rdc-skills 0.8.7 → 0.8.9

package/guides/agent-bootstrap.md

@@ -183,6 +183,47 @@ If you discover that fixing your assigned task would also require changing files
 
 ---
 
+## ⛔ Implementation Report Contract
+
+Every agent MUST follow this protocol before closing any work item as `done`.
+
+### Step 1 — Tick checklist items as you complete them
+
+```sql
+SELECT update_checklist_item('<work-item-id>'::uuid, 'item-id', true);
+```
+
+Call this for each item AS you complete it — not all at once at the end.
+
+### Step 2 — Submit implementation report BEFORE marking done
+
+```sql
+SELECT submit_implementation_report(
+  '<work-item-id>'::uuid,
+  '{"tldr":"...","assumptions":[],"deviations":[],"uncertainty":[],"detail":"...","flags":[]}'::jsonb
+);
+```
+
+Returns `{ flags_count, deviations_count, has_deviations }`. Include this signal in your `AGENT_COMPLETE` report.
+
+### Step 3 — Mark done
+
+```sql
+SELECT update_work_item_status('<work-item-id>'::uuid, 'done');
+```
+
+If any `required: true` checklist item is still unchecked, the DB raises EXCEPTION — fix it first.
+
+### Supervisor workflow
+
+- All zeros → clean run, proceed
+- `flags_count > 0` or `deviations_count > 0` → pull full report:
+  ```sql
+  SELECT implementation_report FROM work_items WHERE id = '<id>';
+  ```
+
+---
+
 ## Now read your role-specific guide
 
 Path: `{PROJECT_ROOT}/.rdc/guides/<type>.md` (e.g., `frontend.md`, `backend.md`, `data.md`)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lifeaitools/rdc-skills",
-  "version": "0.8.7",
+  "version": "0.8.9",
   "description": "RDC typed-agent dispatch skill suite for Claude Code — plan, build, review, overnight builds",
   "keywords": [
     "claude-code",
@@ -19,7 +19,7 @@
     "type": "plugin",
     "skills": "skills/",
     "guides": "guides/",
-    "version": "0.8.7",
+    "version": "0.8.9",
     "commands": "commands/"
   },
   "scripts": {

package/rules/work-items-rpc.md

@@ -122,6 +122,8 @@ const { data } = await supabase.rpc('insert_work_item', {
 | `p_session_id` | text | NULL | Session UUID (for audit trail) |
 | `p_package` | text | NULL | Package/module name (e.g., '@regen/ui') |
 | `p_source` | text | `'manual'` | How was this created: `manual`, `agent`, `api`, `import` |
+| `p_checklist` | jsonb | `NULL` | Checklist items. If NULL and parent has `definition_of_done`, auto-inherited with all `checked: false` |
+| `p_definition_of_done` | jsonb | `NULL` | DoD template (epics only). Child tasks inherit this as their checklist at insert time |
 
 **Returns:**
 - Full inserted work_item row as JSONB
@@ -171,6 +173,7 @@ const { data } = await supabase.rpc('update_work_item_status', {
 
 **Behavior:**
 - Setting `done` auto-sets `completed_at = now()`
+- Setting `done` raises EXCEPTION if any `required: true` checklist item has `checked: false` — DoD gate
 - Setting `todo`, `in_progress`, `blocked`, or `review` clears `completed_at`
 - `updated_at` always refreshed to current timestamp
 - Raises exception if ID not found
@@ -180,7 +183,76 @@ const { data } = await supabase.rpc('update_work_item_status', {
 
 ---
 
-### 4. get_work_items_by_epic — Fetch epic + children
+### 4. update_checklist_item — Tick off a checklist item
+
+**Supabase SQL:**
+```sql
+SELECT update_checklist_item(
+  'work-item-uuid'::uuid,
+  'tsc-clean',   -- string id of the checklist item
+  true           -- checked state
+);
+```
+
+**Parameters:**
+
+| Name | Type | Required | Notes |
+|------|------|----------|-------|
+| `p_work_item_id` | uuid | yes | the work item ID |
+| `p_item_id` | text | yes | the `id` field of the checklist item |
+| `p_checked` | boolean | yes | `true` = checked, `false` = unchecked |
+
+Call this as each checklist item is completed — not all at once at the end.
+
+**Returns:** updated checklist JSONB array.
+
+---
+
+### 5. submit_implementation_report — File agent completion narrative
+
+MUST be called BEFORE `update_work_item_status('done')`.
+
+**Supabase SQL:**
+```sql
+SELECT submit_implementation_report(
+  'work-item-uuid'::uuid,
+  '{
+    "tldr": "One-sentence summary",
+    "assumptions": [],
+    "deviations": [],
+    "uncertainty": [],
+    "detail": "Full narrative",
+    "flags": []
+  }'::jsonb
+);
+```
+
+**Parameters:**
+
+| Name | Type | Required | Notes |
+|------|------|----------|-------|
+| `p_work_item_id` | uuid | yes | the work item ID |
+| `p_report` | jsonb | yes | implementation report (shape below) |
+
+**Returns:** `{ "flags_count": N, "deviations_count": N, "has_deviations": bool }` — signal only.
+
+**`implementation_report` shape:**
+```json
+{
+  "tldr": "string",
+  "assumptions": ["string", "..."],
+  "deviations": ["string — deviation from plan", "..."],
+  "uncertainty": ["string — anything unclear", "..."],
+  "detail": "string — full narrative",
+  "flags": ["string — supervisor attention needed", "..."]
+}
+```
+
+If `has_deviations` is true or `flags_count > 0`, supervisor pulls full report from DB.
+
+---
+
+### 6. get_work_items_by_epic — Fetch epic + children
 
 Get an epic and all its child tasks/subtasks, sorted by priority.
 
@@ -216,7 +288,7 @@ const { data } = await supabase.rpc('get_work_items_by_epic', {
 
 ---
 
-### 5. bump_epic_version — Version with history
+### 7. bump_epic_version — Version with history
 
 Snapshot an epic's current state and assign a version number.
 

package/skills/fixit/SKILL.md

@@ -89,7 +89,13 @@ fi
 
 ### 6. Close and clean up
 
+Submit implementation report first, then mark done:
+
 ```sql
+SELECT submit_implementation_report('<id>'::uuid,
+  '{"tldr":"<one sentence>","assumptions":[],"deviations":[],"uncertainty":[],"detail":"<what was fixed>","flags":[]}'::jsonb
+);
+
 SELECT update_work_item_status('<id>'::uuid, 'done',
   '["Fixed via rdc:fixit"]'::jsonb
 );

package/skills/help/SKILL.md

@@ -4,21 +4,11 @@ description: >-
   Usage `rdc:help` or `rdc` — selection menu of all rdc:* skills with their full argument syntax. Use when unsure which command to invoke or what args it takes.
 ---
 
-> **⚠️ OUTPUT CONTRACT (READ FIRST):** `guides/output-contract.md`
-> Checklist-only output. No tool-call narration. No raw MCP/JSON/log dumps.
-> One checklist upfront, updated in place, shown again at end with a 1-line verdict.
-
 # rdc:help — Command Reference
-> **rdc-skills v0.8.0** — installed at `~/.claude/skills/user/rdc-*.md`
-> To check installed version: `head -3 ~/.claude/skills/user/rdc-help.md`
-> Source: `C:/Dev/rdc-skills` · package.json version field is authoritative
-
-## When to Use
-- Project lead asks "what commands do I have", "what's available", "which skill should I use"
-- You are unsure which `rdc:*` skill to invoke for a given task
-- First session on a project — orient yourself to available commands
 
-Print the full usage menu below verbatim, then ask the project lead which command to run.
+> **⚠️ HARD OUTPUT RULE:** Your ENTIRE response MUST be the block below, copied verbatim.
+> No preamble. No follow-up question. No summary. No "what would you like to do?".
+> Do NOT add any text before or after the block. Emit it exactly as written. This is non-negotiable.
 
 ## Arguments
 

package/skills/plan/SKILL.md

@@ -66,8 +66,9 @@ description: >-
    ```
 
 6. **Create Supabase epic + child tasks:**
-   - Epic via `insert_work_item(p_item_type := 'epic', ...)`
-   - One task per work package via `insert_work_item(p_parent_id := <epic_id>, ...)`
+   - Epic via `insert_work_item(p_item_type := 'epic', p_definition_of_done := '[{"id":"...", "text":"...", "required":true, "checked":false}]'::jsonb, ...)`
+   - Set `p_definition_of_done` on the epic — child tasks inserted under it will auto-inherit it as their checklist
+   - One task per work package via `insert_work_item(p_parent_id := <epic_id>, ...)` — checklist auto-hydrated from epic's DoD
    - Set priorities: urgent/high/normal based on sequencing
 
 7. **Report results:**