opencodekit 0.20.6 → 0.20.8

package/dist/template/.opencode/command/curate.md

@@ -0,0 +1,299 @@
+---
+description: Organize, deduplicate, and curate knowledge in project memory
+argument-hint: "[--scope recent|all] [--auto-merge]"
+agent: build
+---
+
+# Curate: $ARGUMENTS
+
+Organize accumulated knowledge. Surface conflicts, merge duplicates, archive stale observations.
+
+> **Workflow:** `/ship` → `/compound` → **`/curate`** → `/pr`
+>
+> Run periodically (weekly or after major work) to keep memory sharp. Inspired by ByteRover's structured curation pipeline.
+
+## Load Skills
+
+```typescript
+skill({ name: "memory-system" });
+skill({ name: "verification-before-completion" });
+```
+
+## Parse Arguments
+
+| Argument       | Default  | Description                                      |
+| -------------- | -------- | ------------------------------------------------ |
+| `--scope`      | `recent` | `recent` = last 30 days, `all` = entire memory   |
+| `--auto-merge` | false    | Auto-merge exact duplicates without confirmation |
+
+## Phase 1: Inventory
+
+Take stock of current memory state:
+
+```typescript
+memory_admin({ operation: "status" });
+memory_admin({ operation: "capture-stats" });
+```
+
+Report:
+
+```
+## Memory Inventory
+
+| Metric | Count |
+|--------|-------|
+| Total observations | [N] |
+| Recent (30 days) | [N] |
+| By type | pattern: N, decision: N, bugfix: N, ... |
+| Confidence distribution | high: N, medium: N, low: N |
+```
+
+## Phase 2: Domain Detection
+
+Analyze observations to extract semantic domains — groups of related knowledge.
+
+```typescript
+// Get memory status for inventory
+memory_admin({ operation: "status" });
+
+// Search by common concept categories to build domain map
+const domains = [];
+for (const concept of ["build", "test", "memory", "git", "agent", "auth", "ui", "config"]) {
+  const results = memory_search({ query: concept, limit: 20 });
+  // Group results by concept affinity
+}
+```
+
+Categorize observations into domains based on their `concepts` and `title` fields:
+
+| Domain         | Example Concepts                   | Observation Count |
+| -------------- | ---------------------------------- | ----------------- |
+| `build_system` | build, tsdown, rsync, dist         | [N]               |
+| `testing`      | vitest, test, TDD, coverage        | [N]               |
+| `memory`       | observation, FTS5, sqlite, handoff | [N]               |
+| `git_workflow` | commit, branch, push, PR           | [N]               |
+| `agent_system` | subagent, delegation, skills       | [N]               |
+
+**Domain naming rules:**
+
+- snake_case, 1-3 words
+- Semantically meaningful (not just "misc")
+- Maximum 10 domains (merge small groups)
+
+## Phase 3: Conflict & Duplicate Detection
+
+### 3a. Exact Duplicates
+
+```typescript
+memory_admin({ operation: "lint" });
+```
+
+Flag observations with identical or near-identical titles and narratives. Present for merge:
+
+```
+### Duplicates Found
+
+| Obs A | Obs B | Similarity | Recommended Action |
+|-------|-------|------------|-------------------|
+| #12 "Use JWT for auth" | #45 "JWT chosen for auth" | 95% title match | MERGE → keep #45 (newer) |
+| #8 "Build copies .opencode/" | #33 "Build copies .opencode/" | 100% title match | MERGE → keep #33 (newer) |
+```
+
+### 3b. Contradictions
+
+Search for observations where:
+
+- Same concepts but different decisions
+- Same file paths but conflicting patterns
+- Confidence downgrade without supersedes link
+
+```
+### Contradictions Found
+
+| Obs A | Obs B | Conflict | Recommended Action |
+|-------|-------|----------|-------------------|
+| #5 "Always use X pattern" | #29 "Avoid X pattern" | Opposite recommendations | RESOLVE — ask user which is current |
+```
+
+### 3c. Stale Observations
+
+Flag observations where:
+
+- Referenced files no longer exist
+- Referenced patterns no longer appear in codebase
+- Over 90 days old with no related recent activity
+
+```
+### Stale Observations
+
+| Obs | Age | Reason | Recommended Action |
+|-----|-----|--------|-------------------|
+| #3 "src/old-file.ts pattern" | 120 days | File deleted | ARCHIVE |
+| #7 "Use moment.js for dates" | 95 days | Dependency removed | ARCHIVE |
+```
+
+## Phase 4: Present Curation Plan
+
+Compile all findings into a review table:
+
+```
+## Curation Plan
+
+### Actions Required
+
+| # | Observation | Action | Reason |
+|---|------------|--------|--------|
+| 1 | #12 + #45 | MERGE | Duplicate — keep newer |
+| 2 | #5 vs #29 | RESOLVE | Contradicting patterns |
+| 3 | #3 | ARCHIVE | Referenced file deleted |
+| 4 | #7 | ARCHIVE | Dependency removed |
+| 5 | #18 | UPDATE | Low confidence → verify |
+```
+
+```typescript
+question({
+  questions: [
+    {
+      header: "Curation Plan",
+      question: "Review the curation plan. Proceed with all actions?",
+      options: [
+        { label: "Execute all (Recommended)", description: "Apply all actions above" },
+        { label: "Let me cherry-pick", description: "I'll approve individually" },
+        { label: "Skip curation", description: "No changes to memory" },
+      ],
+    },
+  ],
+});
+```
+
+## Phase 5: Execute Curation
+
+For each approved action:
+
+### MERGE (duplicates)
+
+```typescript
+// Read both observations
+const older = memory_get({ ids: "<older-id>" });
+const newer = memory_get({ ids: "<newer-id>" });
+
+// Union-merge: combine comma-separated lists, deduplicate (case-insensitive), existing items first
+// Example: older.facts="auth, jwt" + newer.facts="jwt, session" → "auth, jwt, session"
+
+// Create merged observation (newer as base, merge fields from older)
+observation({
+  type: newer.type,
+  title: newer.title,
+  narrative: newer.narrative,
+  // Manually combine comma-separated fields: keep all unique items from both
+  facts: "[combined unique facts from older + newer]",
+  concepts: "[combined unique concepts from older + newer]",
+  files_modified: "[combined unique file paths from older + newer]",
+  confidence: newer.confidence, // Newer confidence wins
+  supersedes: "<older-id>",
+  subtitle: "Merged from #<older-id> + #<newer-id>",
+});
+```
+
+**Union merge rule:** Combine comma-separated lists, deduplicate (case-insensitive), existing items first.
+
+### RESOLVE (contradictions)
+
+Present the conflicting observations side-by-side:
+
+```
+### Contradiction: #5 vs #29
+
+**#5 (older, high confidence):**
+> Always use X pattern for Y components
+
+**#29 (newer, medium confidence):**
+> Avoid X pattern — causes Z issues
+
+Which is the current truth?
+```
+
+```typescript
+question({
+  questions: [
+    {
+      header: "Resolve Conflict",
+      question: "Which observation reflects the current codebase reality?",
+      options: [
+        { label: "#5 (older) is correct", description: "Archive #29, keep #5" },
+        { label: "#29 (newer) is correct", description: "Supersede #5 with #29" },
+        { label: "Both partially correct", description: "I'll write a reconciled version" },
+      ],
+    },
+  ],
+});
+```
+
+### ARCHIVE (stale)
+
+```typescript
+// Verify staleness by checking codebase
+// If file doesn't exist or pattern not found:
+observation({
+  type: "warning",
+  title: "Archived: [original title]",
+  narrative: "Archived during curation — [reason]. Original observation #<id>.",
+  supersedes: "<stale-id>",
+  confidence: "low",
+});
+```
+
+### UPDATE (low confidence → verify)
+
+```typescript
+// Search codebase for evidence
+// If evidence found → upgrade confidence
+// If evidence not found → archive
+```
+
+## Phase 6: Compile Knowledge Index
+
+After curation, regenerate the knowledge index:
+
+```typescript
+memory_admin({ operation: "compile" });
+memory_admin({ operation: "index" });
+```
+
+## Phase 7: Report
+
+```
+## Curation Summary
+
+**Scope:** [recent / all]
+**Observations reviewed:** [N]
+**Domains identified:** [N]
+
+| Action | Count | Details |
+|--------|-------|---------|
+| Merged | [N] | [list merged pairs] |
+| Resolved | [N] | [list resolved conflicts] |
+| Archived | [N] | [list archived observations] |
+| Updated | [N] | [list confidence changes] |
+| No change | [N] | |
+
+**Memory health:** [Healthy / Needs attention: describe]
+**Next recommended:** /pr or continue work
+```
+
+## When Nothing to Curate
+
+If all observations are clean, well-organized, and non-conflicting:
+
+> "Memory is clean. No duplicates, contradictions, or stale observations found. [N] observations across [M] domains."
+
+Don't force curation. Quality memory means less curation needed.
+
+## Related Commands
+
+| Need                    | Command                        |
+| ----------------------- | ------------------------------ |
+| Extract learnings first | `/compound`                    |
+| Full chain              | `/lfg`                         |
+| Check memory health     | `/health`                      |
+| Search memory           | Use `memory-search()` directly |

package/dist/template/.opencode/command/lfg.md

@@ -61,6 +61,7 @@ Execute the plan:
 
 ```typescript
 skill({ name: "executing-plans" });
+skill({ name: "reflection-checkpoints" }); // Phase transition + mid-point checks
 // Load plan.md, execute wave-by-wave
 // Per-task commits after each task passes verification
 ```

package/dist/template/.opencode/command/ship.md

@@ -19,6 +19,7 @@ skill({ name: "beads" });
 skill({ name: "memory-grounding" });
 skill({ name: "workspace-setup" });
 skill({ name: "verification-before-completion" });
+skill({ name: "reflection-checkpoints" }); // Mid-point + completion checks during execution
 ```
 
 ## Determine Input Type

package/dist/template/.opencode/opencode.json

@@ -109,11 +109,6 @@
 			"enabled": true,
 			"timeout": 120000,
 			"type": "local"
-		},
-		"pencil": {
-			"command": ["openpencil-mcp"],
-			"enabled": false,
-			"type": "local"
 		}
 	},
 	"model": "opencode/minimax-m2.5-free",

package/dist/template/.opencode/package.json

@@ -12,7 +12,7 @@
   },
   "dependencies": {
     "@google/stitch-sdk": "^0.0.3",
-    "@opencode-ai/plugin": "1.4.0"
+    "@opencode-ai/plugin": "1.4.3"
   },
   "devDependencies": {
     "@types/node": "^25.3.0",