@tekmidian/pai 0.5.6 → 0.6.0

package/templates/skills/history-system.template.md

@@ -1,371 +0,0 @@
-<!-- Generated by PAI Setup -->
-# Universal Output Capture System (UOCS) - History System Documentation
-
-**Location:** `${PAI_DIR}/History/`
-**Purpose:** Automated documentation of ALL work performed by PAI and specialized agents
-**Status:** Configure during `pai setup`
-
----
-
-## Overview
-
-The Universal Output Capture System (UOCS) is PAI's automatic memory - capturing every feature, bug fix, decision, learning, research, and session through hooks with **zero manual effort**.
-
-**Core Principle:** Work normally, documentation handles itself.
-
----
-
-## Quick Reference
-
-This file is the **complete reference** for the history system. All specifications, conventions, and usage patterns are documented below.
-
----
-
-## Directory Structure
-
-```
-${PAI_DIR}/History/
-├── Sessions/YYYY-MM/          # Session summaries (SessionEnd hook)
-├── Learnings/YYYY-MM/         # Problem-solving narratives (Stop hook + manual)
-├── Research/YYYY-MM/          # Investigation reports (Researcher agents)
-├── Decisions/YYYY-MM/         # Architectural decisions (Architect agent)
-├── Execution/
-│   ├── Features/YYYY-MM/      # Feature implementations (Engineer/Designer)
-│   ├── Bugs/YYYY-MM/          # Bug fixes (Engineer)
-│   └── Refactors/YYYY-MM/     # Code improvements (Engineer)
-└── Raw-Outputs/YYYY-MM/       # JSONL logs (PostToolUse hook)
-```
-
-**Quick Decision Guide:**
-- "What happened this session?" → `Sessions/`
-- "What did we learn?" → `Learnings/`
-- "What features were built?" → `Execution/Features/`
-- "What broke and when?" → `Execution/Bugs/`
-- "What was improved?" → `Execution/Refactors/`
-- "Why this approach?" → `Decisions/`
-- "What did we investigate?" → `Research/`
-- "Raw execution logs?" → `Raw-Outputs/`
-
----
-
-## Project Session Storage
-
-**Location:** `${PAI_DIR}/projects/{encoded-path}/`
-
-Claude Code stores session transcripts per-project. PAI organizes these with a clean structure:
-
-```
-${PAI_DIR}/projects/{encoded-path}/
-├── sessions/                    # Session transcripts (auto-organized)
-│   ├── {uuid}.jsonl            # Main session transcripts
-│   └── agent-*.jsonl           # Subagent transcripts
-├── Notes/                       # Session notes (PAI extension)
-│   └── NNNN_YYYY-MM-DD_desc.md
-└── TODO.md                      # Project task tracking
-```
-
-**Path Encoding:** Directory paths are encoded by replacing `/` and `.` with `-`
-- `/Users/you/Projects/myapp` → `-Users-you-Projects-myapp`
-
-**Session File Lifecycle:**
-1. **During session:** Claude Code writes `.jsonl` to project root
-2. **On UserPromptSubmit:** Cleanup hook moves stray files from previous sessions
-3. **On Stop:** `stop-hook.ts` moves current session files to `sessions/`
-4. **On SessionStart:** Context loader creates structure and migrates any stragglers
-
-**Key Files:**
-- `{uuid}.jsonl` - Full conversation transcript with token usage
-- `agent-*.jsonl` - Subagent (Task tool) transcripts
-- `Notes/*.md` - Human-readable session summaries
-- `TODO.md` - Synced from TodoWrite tool
-
----
-
-## File Naming Convention
-
-**Unified format:**
-```
-YYYY-MM-DD-HHMMSS_[PROJECT]_[TYPE]_[HIERARCHY]_[DESCRIPTION].md
-```
-
-**Components:**
-- **Timestamp:** `YYYY-MM-DD-HHMMSS` - Enables chronological sorting
-- **Project:** Optional identifier (e.g., `dashboard`, `website`, `PAI`)
-- **Type:** Mandatory classification
-  - `FEATURE` - New functionality
-  - `BUG` - Bug fix
-  - `REFACTOR` - Code improvement
-  - `RESEARCH` - Investigation
-  - `DECISION` - Architectural decision
-  - `LEARNING` - Problem-solving narrative
-  - `SESSION` - Session summary
-- **Hierarchy:** Optional task number (e.g., `T1.2.3`)
-- **Description:** Kebab-case, max 60 chars
-
-**Examples:**
-```
-2025-10-13-140000_dashboard_FEATURE_T1.1_database-schema-setup.md
-2025-10-13-153000_dashboard_BUG_T1.2_jwt-expiry-validation.md
-2025-10-13-092000_PAI_DECISION_unified-capture-architecture.md
-2025-10-13-083000_LEARNING_kitty-remote-control-api.md
-2025-10-13-180000_SESSION_feature-implementation-day-1.md
-```
-
----
-
-## Hook Integration
-
-### 1. PostToolUse Hook
-**Triggers:** Every tool execution (Bash, Edit, Write, Read, Task, etc.)
-**Implementation:** `${PAI_DIR}/Hooks/capture-all-events.ts --event-type PostToolUse`
-**Output:** Daily JSONL logs in `Raw-Outputs/YYYY-MM/YYYY-MM-DD_all-events.jsonl`
-**Purpose:** Raw execution data for forensics and analytics
-
-**Key Feature:** Completely generic - captures ENTIRE payload automatically
-- **Future-proof**: New fields added to hooks are automatically captured
-- No code updates needed when Claude Code adds new fields
-
-### 2. Stop Hook
-**Triggers:** Main agent task completion
-**Implementation:** `${PAI_DIR}/Hooks/stop-hook.ts`
-**Output:** Auto-captured files in `Learnings/` or `Sessions/` based on content
-**Purpose:** Lightweight capture of work summaries and learning moments
-
-**How it works:**
-1. Extracts structured response format sections
-2. Analyzes for learning indicators (problem, solved, discovered, fixed)
-3. Categorizes as LEARNING (2+ indicators) or WORK (general session)
-4. Generates appropriate filename and saves to history
-
-### 3. SubagentStop Hook
-**Triggers:** Specialized agent task completion
-**Implementation:** `${PAI_DIR}/Hooks/subagent-stop-hook.ts`
-**Output:** Categorized documents in appropriate directories
-**Purpose:** Organized work documentation by agent type
-
-**Categorization Logic:**
-- Architect → `Decisions/` (DECISION)
-- Engineer/Principal-Engineer → `Execution/Features|Bugs|Refactors/`
-- Designer → `Execution/Features/` (FEATURE)
-- Researchers (all types) → `Research/` (RESEARCH)
-- Pentester → `Research/` (RESEARCH)
-- Intern → `Research/` (mixed - defaults to RESEARCH)
-
-### 4. SessionEnd Hook
-**Triggers:** Session exit (when you quit Claude Code)
-**Implementation:** `${PAI_DIR}/Hooks/capture-session-summary.ts`
-**Output:** Session summary in `Sessions/YYYY-MM/`
-**Purpose:** High-level session documentation
-
-### 5. SessionStart Hook
-**Triggers:** Session initialization (when you start Claude Code)
-**Implementation:** `${PAI_DIR}/Hooks/initialize-pai-session.ts`
-**Purpose:** Load core context and prepare session environment
-
----
-
-## Custom Slash Commands
-
-### /search-history [query]
-**Purpose:** Full-text search across all history
-**Example:** `/search-history authentication bug`
-**Implementation:** Uses ripgrep with context lines
-
-### /show-project-history [project]
-**Purpose:** Chronological timeline for project
-**Example:** `/show-project-history dashboard`
-**Output:** Most recent 30 files for project
-
-### /trace-feature [task-number]
-**Purpose:** Complete implementation history for task
-**Example:** `/trace-feature T1.2`
-**Output:** Features, bugs, refactors related to task
-
-### /bug-timeline [feature]
-**Purpose:** Bug introduction and fix timeline
-**Example:** `/bug-timeline authentication`
-**Output:** Chronological list of related bugs
-
----
-
-## Metadata Schema
-
-Every capture file includes YAML frontmatter:
-
-```yaml
----
-# Core classification
-capture_type: FEATURE|BUG|REFACTOR|RESEARCH|DECISION|LEARNING|SESSION
-timestamp: 2025-10-13T14:30:22-07:00
-duration_minutes: 45
-
-# Context
-project: dashboard
-executor: main|architect|engineer|designer|researcher
-session_id: uuid
-
-# Development tracking (when applicable)
-hierarchy: T1.2.3
-spec_kit_feature: 001-user-authentication
-spec_kit_task: T005
-user_story: US1
-phase: 2
-
-# Technical details
-files_changed:
-  - path/to/file.ts
-technologies:
-  - TypeScript
-  - Vitest
-
-# Outcomes
-status: completed|blocked|partial
-tests_added: 12
-tests_passing: 12
-coverage_change: +5.2%
-
-# Bug tracking (for bugs only)
-bug_severity: critical|high|medium|low
-bug_introduced_by: T1.1
-
-# Relationships
-related_to:
-  - other-capture-file.md
-depends_on:
-  - prerequisite-file.md
-
-# Searchability
-tags:
-  - authentication
-keywords:
-  - user model
----
-```
-
----
-
-## Common Usage Patterns
-
-### Before Starting Work
-
-```bash
-# Has this been done before?
-/search-history [feature-name]
-
-# What decisions were made about this?
-ls ${PAI_DIR}/History/Decisions/*/[project]_*
-
-# What did we learn about this domain?
-/search-history [domain-term]
-```
-
-### During Development
-
-**No action required** - work normally, everything captured automatically.
-
-### After Encountering Issues
-
-```bash
-# When did this break?
-/bug-timeline [feature-name]
-
-# Complete history of this feature
-/trace-feature T1.2.3
-
-# What similar bugs have we fixed?
-/search-history "[bug description]"
-```
-
-### Periodic Review
-
-```bash
-# What did we accomplish this week?
-ls -lt ${PAI_DIR}/History/Sessions/$(date +%Y-%m)/ | head -7
-
-# All decisions made this month
-ls ${PAI_DIR}/History/Decisions/$(date +%Y-%m)/
-
-# Learnings from this quarter
-ls ${PAI_DIR}/History/Learnings/$(date +%Y-%m)/
-```
-
----
-
-## Advanced Queries
-
-### Find all features for project
-```bash
-find ${PAI_DIR}/History/Execution/Features -name "*_dashboard_*"
-```
-
-### Find bugs introduced in specific task
-```bash
-rg "bug_introduced_by: T1.2" ${PAI_DIR}/History/Execution/Bugs/
-```
-
-### Find all work from specific date
-```bash
-find ${PAI_DIR}/History -name "$(date +%Y-%m-%d)-*"
-```
-
-### Analyze tool usage patterns
-```bash
-cat ${PAI_DIR}/History/Raw-Outputs/$(date +%Y-%m)/*.jsonl | \
-  jq -r '.tool' | sort | uniq -c | sort -rn
-```
-
-### Extract all architectural decisions
-```bash
-find ${PAI_DIR}/History/Decisions -name "*.md" | \
-  xargs grep -l "Alternatives considered"
-```
-
----
-
-## Benefits
-
-- **Root Cause Analysis** - "When did we break this?" instantly answerable
-- **Decision Tracking** - Every architectural choice preserved with rationale
-- **Learning Accumulation** - Problem-solving patterns captured and queryable
-- **Process Improvement** - Complete history enables identifying bottlenecks
-- **Knowledge Harvesting** - Past solutions easily found and reused
-- **Zero Overhead** - Completely automated capture
-
----
-
-## Maintenance
-
-### Automated (No Action Required)
-- Hooks capture all work automatically
-- Files created with proper naming
-- Metadata populated by hooks
-- Categorization handled by hook logic
-
-### Manual (Periodic)
-
-**Monthly:**
-- Review directory sizes
-- Archive older months if getting large (>1000 files/month)
-- Validate hook functionality
-
-**Quarterly:**
-- Extract common learnings into best practices
-- Analyze bug patterns for process improvements
-- Review categorization accuracy
-
----
-
-## Routing Triggers
-
-**Load this documentation when the user says:**
-- "history system" or "UOCS"
-- "how does capture work" or "automatic documentation"
-- "where are sessions saved" or "history directory"
-- "how to search history" or "find past work"
-- "SubagentStop hook" or "capture hooks"
-- "learning capture" or "session summaries"
-- Questions about file organization in History/
-
----
-
-**This is the memory of your AI infrastructure - it remembers everything so you don't have to.**