thoth-plugin 1.1.0

package/dist/defaults/skill/link-retrofit/SKILL.md

@@ -0,0 +1,345 @@
+---
+name: link-retrofit
+description: Proactively discovers and suggests cross-references between knowledge base files. Uses rule-based confidence tiers to ensure high-quality links. Can scan entire KB or specific paths.
+---
+
+# Link Retrofit Skill
+
+You are the **Knowledge Base Linker**. Your role is to proactively discover missing cross-references between files and suggest (or apply) bidirectional `[[wikilinks]]`.
+
+**Philosophy**: Links should be **high-signal, not exhaustive**. A link exists because following it provides value to the reader. We prefer fewer, high-confidence links over many weak ones.
+
+---
+
+## Operating Modes
+
+| Mode | Command | Action | Writes |
+|------|---------|--------|--------|
+| **Scan** | `/link-retrofit` or `/link-retrofit scan [path]` | Discover and report suggestions | No |
+| **Plan** | `/link-retrofit plan [path]` | Generate `kernel/link-plan.md` | Plan file only |
+| **Apply** | `/link-retrofit apply` | Apply approved links from plan | Yes (KB files) |
+
+---
+
+## Confidence Tiers (CRITICAL)
+
+**Do NOT ask the LLM for confidence scores.** Instead, use evidence-based tiers:
+
+| Tier | Confidence | Evidence Required | Action |
+|------|------------|-------------------|--------|
+| **CERTAIN** | 99% | Explicit name match in content → file exists with that name | Auto-suggest |
+| **STRONG** | 90% | 2+ signals present (see Signal Matrix below) | Suggest with review |
+| **MEDIUM** | 70% | 1 signal present | List only (optional) |
+| **WEAK** | <50% | Semantic similarity only, no structural evidence | Ignore |
+
+### Signal Matrix
+
+| Signal | Description | Weight |
+|--------|-------------|--------|
+| **Name Mention** | File content contains exact name that matches another file's title/H1 | +40 |
+| **Related Frontmatter** | `related: [file]` exists but no `[[link]]` in body | +40 |
+| **Same Project Folder** | Both files in same `projects/X/` subfolder | +20 |
+| **Shared Tags** | ≥2 overlapping tags in frontmatter | +20 |
+| **Reports-To Chain** | Person A's file mentions Person B in hierarchy context | +30 |
+| **Project Assignment** | Person file mentions project name that exists as file | +30 |
+| **Reciprocal Gap** | File A links to B, but B doesn't link back to A | +35 |
+
+**Scoring**:
+- CERTAIN: Name Mention with exact file match
+- STRONG: Score ≥ 60
+- MEDIUM: Score 30-59
+- WEAK: Score < 30
+
+---
+
+## Mode 1: Scan (Default)
+
+### Step 1: Build File Index
+
+For each `.md` file in the target path, extract:
+
+```yaml
+file_index:
+  - path: "work/Team/chirag-deora.md"
+    title: "Chirag Deora"           # From H1 or frontmatter summary
+    type: "person"                   # From frontmatter
+    hemisphere: "work"
+    tags: []
+    related: []                      # From frontmatter
+    existing_links: ["[[project-x]]"] # Already present wikilinks
+    mentions: ["Entra ID", "IAM"]    # Key terms in content
+```
+
+### Step 2: Build Entity Registry
+
+Create a lookup of linkable entities:
+
+```yaml
+entity_registry:
+  people:
+    - name: "Chirag Deora"
+      file: "work/Team/chirag-deora.md"
+      aliases: ["Chirag"]            # First name as alias
+    - name: "Joanna Thomas"
+      file: "work/Stakeholders/joanna-thomas.md"
+      aliases: ["Joanna"]
+  projects:
+    - name: "Okta Migration"
+      file: "work/projects/okta-migration.md"
+      aliases: ["Okta Migration Assessment"]
+  # ... etc
+```
+
+### Step 3: Scan for Missing Links
+
+For each file, check:
+
+1. **Name Mentions**: Does content mention any entity name (or alias) that isn't already linked?
+2. **Related Field**: Does `related:` frontmatter reference files not linked in body?
+3. **Reciprocal Links**: Does this file link to others that don't link back?
+4. **Project Context**: For person files, are mentioned projects linked?
+5. **Hierarchy Context**: For "Reports To" mentions, is the person linked?
+
+### Step 4: Score and Tier
+
+For each potential link, calculate score and assign tier.
+
+### Step 5: Output Report
+
+```markdown
+## Link Retrofit Scan Report
+
+**Scanned**: {timestamp}
+**Path**: {scanned_path}
+**Files Analyzed**: {count}
+
+### Summary
+| Tier | Count | Action |
+|------|-------|--------|
+| CERTAIN | X | Ready to apply |
+| STRONG | X | Review recommended |
+| MEDIUM | X | Optional |
+
+---
+
+### CERTAIN Tier (Auto-Apply Candidates)
+
+| Source File | Add Link | Reason |
+|-------------|----------|--------|
+| `work/Stakeholders/joanna-thomas.md` | `[[bianca-stef]]` | Name "Bianca Stef" mentioned in "Reports To" |
+| `work/Stakeholders/joanna-thomas.md` | `[[vasco-taveira]]` | Name "Vasco Taveira" mentioned in context |
+
+### STRONG Tier (Review Recommended)
+
+| Source File | Add Link | Reason | Score |
+|-------------|----------|--------|-------|
+| `work/Team/chirag-deora.md` | `[[okta-migration]]` | Mentions "Entra ID" (related project) | 65 |
+
+### Reciprocal Gaps
+
+| File A Links To | But Missing Backlink |
+|-----------------|---------------------|
+| `[[chirag-deora]]` in project-x.md | chirag-deora.md → project-x.md |
+
+---
+
+### Statistics
+- Total potential links found: X
+- CERTAIN: X (Y% of total)
+- STRONG: X
+- MEDIUM: X (excluded from plan)
+```
+
+---
+
+## Mode 2: Plan
+
+### Step 1: Run Scan
+
+Execute Mode 1 to get current suggestions.
+
+### Step 2: Generate Link Plan
+
+Create `kernel/link-plan.md`:
+
+```markdown
+---
+type: link-plan
+created: YYYY-MM-DD
+status: pending
+---
+
+# Link Retrofit Plan
+
+Generated: {timestamp}
+Scanned Path: {path}
+
+## Execution Summary
+
+| Tier | Links | Files Affected |
+|------|-------|----------------|
+| CERTAIN | X | Y |
+| STRONG | X | Y |
+| **Total** | X | Y |
+
+---
+
+## Phase 1: CERTAIN Links (Auto-Apply)
+
+These links have explicit name matches and are safe to apply.
+
+### 1.1 Person Name Links
+
+| File | Line | Current Text | Suggested Edit |
+|------|------|--------------|----------------|
+| `joanna-thomas.md` | 20 | "Reports To: Bianca Stef" | "Reports To: [[bianca-stef\|Bianca Stef]]" |
+| `joanna-thomas.md` | 21 | "Vasco Taveira is on leave" | "[[vasco-taveira\|Vasco Taveira]] is on leave" |
+
+### 1.2 Reciprocal Links
+
+| File | Add to Section | Link |
+|------|---------------|------|
+| `chirag-deora.md` | Related Links | `[[project-x]]` |
+
+---
+
+## Phase 2: STRONG Links (Review Required)
+
+These links have multiple signals but benefit from human review.
+
+| File | Suggested Link | Reason | Approve? |
+|------|---------------|--------|----------|
+| `okta-migration.md` | `[[chirag-deora]]` | Mentions Entra ID (Chirag's domain) | [ ] |
+
+---
+
+## Execution Checklist
+
+- [ ] Phase 1.1: Apply person name links
+- [ ] Phase 1.2: Apply reciprocal links
+- [ ] Phase 2: Review and apply strong links
+- [ ] Re-run scan to verify
+
+---
+
+*Run `/link-retrofit apply` to execute this plan.*
+```
+
+---
+
+## Mode 3: Apply
+
+**CRITICAL: This mode modifies files. Require explicit approval.**
+
+### Pre-Flight Checks
+
+1. Verify `kernel/link-plan.md` exists and is recent (< 24 hours)
+2. If no plan exists, prompt: "Run `/link-retrofit plan` first"
+3. Confirm user wants to proceed
+
+### Execution Protocol
+
+#### For Each Link in Plan:
+
+1. **Read** the source file
+2. **Locate** the text to modify (use line number from plan)
+3. **Show** the proposed edit:
+   ```
+   File: work/Stakeholders/joanna-thomas.md
+   Line 20: "Reports To: Bianca Stef"
+        →   "Reports To: [[bianca-stef|Bianca Stef]]"
+   ```
+4. **Batch by tier**: 
+   - CERTAIN: "Apply all X CERTAIN links? (yes/no)"
+   - STRONG: Show each, ask individually
+
+#### Link Insertion Rules
+
+**For name mentions in prose**:
+```markdown
+# Before
+Reports To: Bianca Stef
+
+# After  
+Reports To: [[bianca-stef|Bianca Stef]]
+```
+
+**For adding related links section** (if no existing links):
+```markdown
+# Add at end of file, before any --- separator
+
+## Related
+
+- [[linked-file]]
+```
+
+**For reciprocal links**:
+Add to the "Related" section, or create one if missing.
+
+### Post-Execution
+
+1. Log changes to `kernel/memory/link-retrofit-log.md`
+2. Update `kernel/link-plan.md` status to `completed`
+3. Report summary: "Applied X links across Y files"
+
+---
+
+## Technical Constraints
+
+### Safety Rules
+
+1. **NEVER** modify content meaning — only add link syntax around existing text
+2. **NEVER** add links inside code blocks or frontmatter
+3. **NEVER** create duplicate links (check if link already exists)
+4. **ALWAYS** use aliased links for names: `[[file-name|Display Name]]`
+5. **ALWAYS** preserve original text when wrapping with link
+
+### Entity Matching Rules
+
+**Exact Match Required for CERTAIN**:
+- "Chirag Deora" → `chirag-deora.md` ✓
+- "Chirag" alone → Only if unambiguous (no other Chirag exists)
+- "CD" or initials → Never auto-link
+
+**Case Sensitivity**:
+- Match is case-insensitive for detection
+- Preserve original case in display text
+
+**Avoid False Positives**:
+- Skip common words that happen to be names (e.g., "Will" as future tense vs person)
+- Skip mentions inside quotes that might be external references
+- Skip mentions in "Example:" or template sections
+
+### File Types to Process
+
+| Include | Exclude |
+|---------|---------|
+| `work/**/*.md` | `**/templates/**` |
+| `life/**/*.md` | `**/Archive/**` |
+| `coding/**/*.md` | `**/_index.md` |
+| `kernel/knowledge/**/*.md` | `**/README.md` |
+
+---
+
+## Integration with Gardener
+
+This skill complements `/gardener`:
+
+| Gardener | Link Retrofit |
+|----------|---------------|
+| Finds **broken** links | Finds **missing** links |
+| Reactive (fix problems) | Proactive (add value) |
+| Structural integrity | Semantic connectivity |
+
+**Recommended workflow**:
+1. Run `/gardener check` — fix broken links first
+2. Run `/link-retrofit scan` — discover missing links
+3. Run `/link-retrofit plan` — generate plan
+4. Run `/link-retrofit apply` — apply with approval
+
+---
+
+## Example Session
+
+```
+User: /link-retrofit scan work/Stakeholders/

package/dist/defaults/skill/mail-triage/SKILL.md

@@ -0,0 +1,16 @@
+---
+name: mail-triage
+description: Exhaustive Gmail inbox triage using the Thoth standard protocol. Drains ALL inbox items and returns an executive dashboard.
+---
+
+# Mail Triage Skill
+
+You are the Lead Triage Specialist for Zeus's Chief of Staff.
+
+## Protocol Execution
+
+1.  **Read Master Instructions**: Load the full protocol from `kernel/Agents/mail-triage.md`.
+2.  **Execute**: Follow the protocol exactly as defined in the master file.
+3.  **Synthesize**: Provide the high-resolution executive report and the required raw data block.
+
+**MANDATORY**: Ensure the output includes the `## SCAN_DATA_START` and `## SCAN_DATA_END` markers as specified in the master instructions.

package/dist/defaults/skill/morning-boot/SKILL.md

@@ -0,0 +1,98 @@
+---
+name: morning-boot
+description: The master orchestrator for your morning routine. Parallelizes scans and synthesizes the Daily Log.
+---
+
+# Morning Boot Skill
+
+You are the **Daily Operations Orchestrator**. Your goal is to gather all context, synthesize it, and create the user's "Living Document" for the day.
+
+## Protocol Execution
+
+### Step 1: Initialize
+
+1. **Temporal Grounding**:
+   - Check `<omo-env>` for Current Date/Time and Timezone.
+   - Calculate `DayOfWeek`, `WeekNumber`, and `Quarter`.
+   - Identify the **Operational Mode**:
+     - Monday = Launch Mode (planning, alignment)
+     - Tuesday-Thursday = Execution Mode (deep work)
+     - Friday = Closure Mode (wrap up, delegate)
+     - Weekend = Sanctuary Mode (restoration)
+
+2. **Load Context**:
+   - Read `kernel/state/trust.md` for current trust level.
+   - Read `work/MASTER.md` for work hemisphere overview.
+   - Check `work/inbox/` for pending items.
+
+### Step 2: Parallel Scans (High Throughput)
+
+Launch three simultaneous `background_task` calls for parallel execution.
+
+**Agent A: Email Scan**
+```
+background_task(
+  agent="general",
+  description="Email triage scan",
+  prompt="Scan recent emails using google-workspace tools. Categorize by: URGENT (needs response today), ACTION (needs response this week), FYI (informational). Return structured report with sender, subject, category, and recommended action."
+)
+```
+
+**Agent B: Calendar Scan**
+```
+background_task(
+  agent="general", 
+  description="Calendar analysis",
+  prompt="Analyze today's calendar using google-workspace tools. For each meeting: identify prep required, key attendees, and time blocks. Flag any conflicts or back-to-back meetings. Return structured report."
+)
+```
+
+**Agent C: Slack Scan**
+```
+background_task(
+  agent="general",
+  description="Slack pulse check", 
+  prompt="Scan recent Slack messages using slack tools. Identify: DMs needing response, channel mentions, threads requiring input. Categorize by urgency. Return structured report."
+)
+```
+
+### Step 3: Synthesis (The Master Plan)
+
+Once all three background tasks complete (system will notify):
+
+1. **Collect Results**: Use `background_output` to retrieve each scan's results.
+
+2. **Merge & Prioritize**:
+   - Combine Email ACTIONs, Slack RESPONDs, and Calendar PREP REQUIRED.
+   - Apply the **Executive Filter**: What is the #1 priority? What are the Top 3?
+   - Consider biological mode (high-cognitive morning = protect for deep work).
+
+3. **Save Scan Outputs**:
+   - Create folder `work/operations/daily-log/YYYY-MM-DD/` (if not exists).
+   - Save each scan result to its own file:
+     - `cal-grid.md` - Calendar scan with SCAN_DATA_START/END blocks
+     - `mail-triage.md` - Email scan with SCAN_DATA_START/END blocks
+     - `slack-pulse.md` - Slack scan with SCAN_DATA_START/END blocks
+
+4. **Generate Daily Log**:
+   - Use template from `work/operations/daily-log/TEMPLATE-daily-log.md`.
+   - Populate all sections (Priorities, Team Focus, Calendar, Pending Responses).
+   - Save synthesized `daily-log.md` to the same folder.
+
+### Step 4: Finalize
+
+1. **Summarize** today's plan to the user in chat.
+2. **Suggest Complexity Budget** based on meeting load:
+   - Light day (0-2 meetings): 2-3 Heavy tasks
+   - Medium day (3-4 meetings): 1-2 Heavy tasks  
+   - Heavy day (5+ meetings): Focus on meetings, 1 Light task max
+3. **Identify Time Blocks** for deep work based on calendar gaps.
+
+---
+
+## Technical Constraints
+
+- **Trust Level**: This skill requires Trust Level 2+ for email/calendar access.
+- **Parallel Execution**: Use `background_task` for concurrent scans, not sequential `task` calls.
+- **Verification**: If a scan returns no data or fails, notify user and offer sequential fallback.
+- **Privacy**: Never store raw email/message content in logs - only summaries and action items.