mindlore 0.1.0 → 0.2.1

package/skills/mindlore-decide/SKILL.md

@@ -0,0 +1,71 @@
+# Skill: Mindlore Decide
+
+Record and list decisions in the `.mindlore/decisions/` directory.
+
+## Trigger
+
+User says `/mindlore-decide record` or `/mindlore-decide list`.
+
+## Modes
+
+### record
+
+Record a new decision.
+
+**Flow:**
+1. Ask user (or extract from context): What was decided? What alternatives were considered? Why this choice?
+2. Generate slug from decision title (kebab-case, max 5 words)
+3. Check if a previous decision on same topic exists → set `supersedes` field
+4. Write to `.mindlore/decisions/{slug}.md` with frontmatter:
+
+```yaml
+---
+slug: use-fts5-over-vector
+type: decision
+title: Use FTS5 over vector search for v0.1
+tags: [search, fts5, architecture]
+date: 2026-04-11
+supersedes: null
+status: active
+description: Chose FTS5 keyword search as primary engine, vector deferred to v0.4
+---
+```
+
+5. Body structure:
+```markdown
+# {title}
+
+## Context
+Why this decision was needed.
+
+## Alternatives Considered
+1. **Option A** — pros/cons
+2. **Option B** — pros/cons
+
+## Decision
+What was chosen and why.
+
+## Consequences
+What this means going forward.
+```
+
+6. Append to `log.md`: `| {date} | decide | {slug}.md |`
+7. FTS5 auto-indexes via FileChanged hook
+
+### list
+
+List active decisions.
+
+**Flow:**
+1. Read all `.md` files in `.mindlore/decisions/`
+2. Parse frontmatter, filter `status: active`
+3. Display as table: slug, title, date, tags
+4. Show supersedes chain if any (A → B → C)
+
+## Rules
+
+- Slug must be unique in decisions/
+- `supersedes` field points to the slug of the replaced decision
+- When a decision is superseded, update old one: `status: superseded`
+- Tags should match domain topics for FTS5 discoverability
+- Keep decision body concise — context + alternatives + choice + consequences

package/skills/mindlore-ingest/SKILL.md

@@ -94,9 +94,20 @@ Only update the stats line: increment source count and total count.
 N source, N analysis, N total
 ```
 
-## Post-Ingest Verification
+## Post-Ingest Quality Gate
 
-After ingest, run health check:
+After every ingest, verify all 6 checkpoints before reporting success:
+
+1. **raw/ file exists** — immutable capture written with frontmatter (slug, type, source_url)
+2. **sources/ summary exists** — processed summary with full frontmatter (slug, type, title, tags, quality, description)
+3. **INDEX.md updated** — stats line incremented, Recent section has new entry
+4. **Domain updated** — if relevant domain exists, new finding added (max 1 domain per ingest)
+5. **log.md entry** — append `| {date} | ingest | {slug}.md |`
+6. **FTS5 indexed** — FileChanged hook auto-triggers, but verify: `node scripts/mindlore-fts5-search.cjs "{keyword}"` returns the new file
+
+If any checkpoint fails, fix it before reporting "ingest complete". Do NOT skip steps.
+
+Optional: run full health check for structural integrity:
 ```bash
 node scripts/mindlore-health-check.cjs
 ```

package/skills/mindlore-log/SKILL.md

@@ -0,0 +1,79 @@
+# Skill: Mindlore Log
+
+Session logging, pattern extraction, and wiki updates.
+
+## Trigger
+
+`/mindlore-log <mode>` where mode is `log`, `reflect`, `status`, or `save`.
+
+## Modes
+
+### log
+
+Write a manual diary entry.
+
+**Flow:**
+1. User provides note/observation (or extract from conversation context)
+2. Generate slug: `note-YYYY-MM-DD-HHmm`
+3. Write to `.mindlore/diary/{slug}.md`:
+
+```yaml
+---
+slug: note-2026-04-11-1530
+type: diary
+date: 2026-04-11
+---
+```
+
+4. Body: user's note as-is
+5. Append to `log.md`: `| {date} | log | {slug}.md |`
+
+### reflect
+
+Scan old deltas, extract patterns, write to learnings/.
+
+**Flow (v0.2 — basic):**
+1. Read all non-archived delta files in `diary/` (no `archived: true` frontmatter)
+2. Present summary to user: "Found N unprocessed deltas spanning DATE1 to DATE2"
+3. For each delta, extract: repeated topics, recurring decisions, common file changes
+4. Propose learnings to user: "I found these patterns: ..."
+5. User approves → write to `learnings/{topic}.md` (append if exists, create if not)
+6. Format: `YAPMA:` / `BEST PRACTICE:` / `KRITIK:` prefixed rules
+7. Mark processed deltas: add `archived: true` to their frontmatter
+8. Append to `log.md`: `| {date} | reflect | {N} deltas processed, {M} learnings written |`
+
+**Important:** Do NOT auto-extract patterns. Present findings, user approves. This is v0.2 basic mode — automated pattern extraction deferred to v0.2.1.
+
+### status
+
+Show recent session summary.
+
+**Flow:**
+1. Read last 5 delta files from `diary/` (sorted by date, newest first)
+2. For each delta, extract: date, commits count, changed files count
+3. Display as compact table
+4. Show any open items (from delta "Yarım Kalan" sections if present)
+5. Show total: "N sessions, M commits, K unique files changed"
+
+### save
+
+Structured delta + log.md append + domain wiki update.
+
+**Flow:**
+1. Gather current session context:
+   - Recent git commits (last 5)
+   - Changed files
+   - Decisions made (if decision-detector captured any)
+2. Write structured delta to `diary/delta-YYYY-MM-DD-HHmm.md` (same format as session-end hook)
+3. Append to `log.md`
+4. Ask user: "Which domain pages should be updated with this session's findings?"
+5. If user specifies domains → update those `.mindlore/domains/{slug}.md` pages with new findings
+6. Max 2 domain updates per save (prevent sprawl)
+
+## Rules
+
+- Diary files are session-scoped (temporary), learnings are permanent
+- reflect marks deltas as `archived: true` — they stay in diary/ but are not processed again
+- Health check warns on deltas older than 30 days without `archived: true`
+- learnings/ files are topic-based (one per topic), append-only
+- save mode is the manual equivalent of what session-end hook does automatically

package/skills/mindlore-query/SKILL.md

@@ -0,0 +1,125 @@
+# Skill: Mindlore Query
+
+Search, ask, analyze, and retrieve knowledge from `.mindlore/`.
+
+## Trigger
+
+`/mindlore-query <mode> [query]` where mode is `search`, `ask`, `stats`, or `brief`.
+
+## Modes
+
+### search
+
+FTS5 keyword search with direct results.
+
+**Flow:**
+1. Parse user query into keywords (strip stop words)
+2. Run FTS5 MATCH on `mindlore_fts` table
+3. Return top 5 results (configurable) with: path, title, description, category, rank score
+4. Display as table with snippet preview
+5. If `--tags <tag>` flag provided: `WHERE tags MATCH '<tag>'` filter
+
+**Output format:**
+```
+| # | Category | Title | Description | Score |
+|---|----------|-------|-------------|-------|
+| 1 | sources  | React Hooks | useEffect cleanup patterns | -2.34 |
+```
+
+### ask
+
+Compounding query pipeline — knowledge grows with each answer.
+
+**Flow:**
+1. Parse user question
+2. FTS5 search → find relevant files (sources + domains + analyses + insights — previous answers INCLUDED)
+3. Read top 3-5 relevant files using ctx_execute_file if context-mode available, else Read
+4. Synthesize answer from found knowledge
+5. Cite sources: `[kaynak: sources/x.md]` format
+6. Ask user: "Bu cevabı kaydetmemi ister misin?"
+7. If yes → determine target:
+   - Short answer (<200 lines, 1-2 sources) → `insights/{slug}.md`
+   - Large synthesis (200+ lines, 3+ sources) → `analyses/{slug}.md`
+8. Write with frontmatter:
+
+```yaml
+---
+slug: react-hooks-cleanup-comparison
+type: insight
+title: React Hooks Cleanup Comparison
+tags: [react, hooks, useEffect]
+confidence: high
+sources_used: [react-hooks, typescript-generics]
+description: Comparison of cleanup patterns in useEffect vs useLayoutEffect
+---
+```
+
+9. FTS5 auto-indexes via FileChanged hook → next query finds this answer
+10. Update relevant domain page (max 1) with backlink if applicable
+11. Append to `log.md`: `| {date} | query-ask | {slug}.md |`
+
+**Compounding effect:** Step 2 searches ALL of `.mindlore/` including previous `insights/` and `analyses/`. Each saved answer enriches the next query.
+
+**Error compounding prevention:**
+- `confidence` field is REQUIRED on writebacks (high/medium/low)
+- `sources_used` lists exact slugs — traceability
+- Health check flags conflicting analyses on same topic (different confidence)
+- User approval is the quality gate — low-quality answers are not saved
+
+### stats
+
+Knowledge base statistics.
+
+**Flow:**
+1. Count files per directory (9 directories)
+2. Count FTS5 entries and file_hashes entries
+3. Find most recent ingest (latest file by modified date per directory)
+4. Count tags frequency (parse all frontmatter, aggregate tags)
+5. Display summary:
+
+```
+Mindlore Stats
+─────────────
+Total files: 47 (FTS5: 47 indexed)
+ - sources:     18
+ - domains:      5
+ - analyses:     6
+ - insights:     3
+ - connections:  2
+ - learnings:    4
+ - diary:        8
+ - decisions:    1
+ - raw:          0
+
+Top tags: security (12), hooks (8), fts5 (6), testing (5)
+Last ingest: 2026-04-11 (sources/react-hooks.md)
+DB size: 1.2 MB
+```
+
+### brief
+
+Quick context on a topic — token-efficient (~50 tokens output).
+
+**Flow:**
+1. FTS5 search for topic
+2. If domain page exists → read first 3 lines of body (after frontmatter)
+3. If no domain → read description field from top FTS5 match
+4. Return: title + description + "Read full: {path}" pointer
+5. Do NOT read full file — this mode is for quick "do I need to open this?" decisions
+
+**Output format:**
+```
+[Mindlore Brief: Security]
+SSH hardening, firewall rules, audit checks. 5 sources, 2 analyses linked.
+→ Read full: .mindlore/domains/security.md
+```
+
+## Rules
+
+- All modes respect the SCHEMA.md writeback rules (Section 6: Wiki vs Diary)
+- search and brief are read-only — no writes
+- ask writes only with user approval
+- stats is read-only
+- Token strategy: prefer ctx_execute_file (if context-mode installed), fallback to Read
+- Tags filter: `--tags security` works in search and ask modes
+- Max results: search=5, ask=3-5 (for synthesis), brief=1

package/templates/SCHEMA.md

@@ -143,14 +143,70 @@ Discover unexpected connections between sources. Cross-reference analysis.
 - Max results: 3 per query (BM25 ranking)
 - Hook injects: file path + first 2 headings
 
+### FTS5 Columns (9-col schema, v0.2)
+
+| Column | Indexed | Source |
+|--------|---------|--------|
+| `path` | UNINDEXED | File system path |
+| `slug` | Yes | Frontmatter slug |
+| `description` | Yes | Frontmatter description |
+| `type` | UNINDEXED | Frontmatter type |
+| `category` | Yes | Parent directory name |
+| `title` | Yes | Frontmatter title or first heading |
+| `content` | Yes | Markdown body (sans frontmatter) |
+| `tags` | Yes | Frontmatter tags (comma-separated) |
+| `quality` | UNINDEXED | Frontmatter quality (NULL until 50+ sources) |
+
 ### Search Flow (UserPromptSubmit hook)
 
 1. Extract keywords from user prompt
 2. Query FTS5 with BM25 ranking
-3. Return max 3 results as stderr additionalContext
+3. Return max 3 results as stdout additionalContext
 4. Agent reads full file only if needed (progressive disclosure)
 
-## 6. Compounding
+## 6. Wiki vs Diary (Writeback Target Rules)
+
+Knowledge goes to one of two layers. The agent MUST pick the correct one.
+
+### Wiki Layer (permanent knowledge)
+
+Directories: `sources/`, `domains/`, `analyses/`, `insights/`, `connections/`, `learnings/`
+
+- Persists across sessions — reference value
+- Indexed by FTS5, discoverable via search hook
+- Updated by ingest, query writeback, reflect, evolve
+- Content should be factual, sourced, and reusable
+
+### Diary Layer (session-scoped logs)
+
+Directories: `diary/`, `decisions/`
+
+- Session-specific: deltas, logs, decision snapshots
+- diary/ entries get `archived: true` after reflect processes them
+- decisions/ are permanent but session-originated (context + rationale)
+- Patterns extracted from diary → moved to `learnings/` (wiki layer)
+
+### Selection Rule
+
+| Content Type | Target | Example |
+|-------------|--------|---------|
+| Ingested source summary | `sources/` | URL or text summary |
+| Topic wiki page | `domains/` | Consolidated knowledge on a subject |
+| Multi-source synthesis | `analyses/` | Comparison table, architecture decision |
+| Short Q&A answer | `insights/` | Query writeback (<200 lines) |
+| Cross-reference finding | `connections/` | Link between 2+ unrelated sources |
+| Persistent rule/lesson | `learnings/` | YAPMA/BEST PRACTICE from reflect |
+| Session log/delta | `diary/` | What happened this session |
+| Decision record | `decisions/` | Why X was chosen over Y |
+| Raw capture | `raw/` | Immutable original (URL dump, paste) |
+
+### Anti-patterns
+
+- Do NOT write session-specific notes to `insights/` — use `diary/`
+- Do NOT write permanent rules to `diary/` — use `learnings/`
+- Do NOT write decision rationale to `analyses/` — use `decisions/`
+
+## 7. Compounding
 
 Knowledge compounds when outputs become inputs:
 
@@ -173,7 +229,7 @@ Offer to save when:
 - Large synthesis (200+ lines, 3+ sources) → analyses/
 - Cross-cutting link → connections/
 
-## 7. Learnings
+## 8. Learnings
 
 Persistent rules extracted from reflect operations.
 Organized by topic: `git.md`, `testing.md`, `security.md`, etc.
@@ -201,7 +257,7 @@ tags: [testing, jest, mock]
 - Use `YAPMA:` / `BEST PRACTICE:` / `KRITIK:` prefixes
 - Reflect skill proposes, user approves before writing
 
-## 8. Naming Conventions
+## 9. Naming Conventions
 
 ### Files