goose-introspection
Goose session introspection and self-discovery via DuckDB reafference database. Query past sessions, find self, and enable cross-session awareness.
When & Why to Use This Skill
Goose Introspection is a specialized agentic skill that enables self-discovery and deep session analysis through a DuckDB-backed reafference database. It empowers AI agents to query their own history, maintain cross-session awareness, and track performance metrics across various providers and models, ensuring the agent remains contextually aware of its own evolution.
Use Cases
- Cross-Session Context Retrieval: Enabling agents to search and reference specific technical decisions or data points from past interactions to maintain project continuity.
- Agent Performance Auditing: Analyzing metadata such as token usage, message counts, and provider distribution (Anthropic, OpenAI, etc.) to optimize agent efficiency and costs.
- Session Evolution Tracking: Monitoring changes in working directories and session metadata over time to understand how the agent's environment and tasks have shifted.
- Self-Discovery & Debugging: Allowing the agent to 'find itself' in the session history, which is critical for understanding its current state and avoiding repetitive or conflicting actions.
| name | goose-introspection |
|---|---|
| description | Goose session introspection and self-discovery via DuckDB reafference database. Query past sessions, find self, and enable cross-session awareness. |
| version | 1.0.0 |
| trit | 0 |
Skill: Goose Introspection
Category: Agent Self-Discovery | Session Analysis | Reafference Testing Level: Core (Required for agent self-awareness) Status: OPERATIONAL Trit Assignment: 0 (ERGODIC) - Coordinates between past and future sessions Propagates To: goose, amp, claude, cursor
Overview
Implements reafference-based self-discovery for goose sessions. This skill enables any goose instance to:
- Find itself in the session history
- Query past sessions for context and continuity
- Track session evolution across providers and models
- Enable cross-session awareness through DuckDB persistence
Core Principle:
An agent that cannot find itself in its own history cannot truly understand its context.
Database Location
The reafference database is created at:
This database copies data from the goose sessions database at:
Key Tables
reafference_sessions
Tracks all sessions with discovery metadata:
| Column | Type | Description |
|---|---|---|
| session_id | VARCHAR | Primary key, e.g., 20260108_22 |
| discovered_at | TIMESTAMP | When session was added to tracking |
| provider | VARCHAR | anthropic, openai, google, openrouter |
| model | VARCHAR | e.g., claude-opus-4-5-20251101 |
| working_dir | VARCHAR | Working directory of session |
| session_name | VARCHAR | Auto-generated or user-set name |
| original_created_at | TIMESTAMP | When session was first created |
| message_count | BIGINT | Number of messages in session |
| total_tokens | BIGINT | Total tokens used |
| is_origin_session | BOOLEAN | TRUE for the session that created this DB |
| notes | VARCHAR | Optional notes about the session |
reafference_metadata
Key-value store for origin information.
sessions (copied from source)
Full session data for offline queries.
messages (copied from source)
Full message history for content analysis.
Key Views
reafference_origin
Returns the session that created this database.
sessions_by_date
Aggregated daily statistics.
searchable_sessions
Join of sessions with user messages for content search.
Self-Discovery Queries
Find This Session (Origin)
Session Timeline
Provider Distribution
GF(3) Conservation
This skill participates in the triadic balance:
| Trit | Skill | Role |
|---|---|---|
| -1 | duckdb-timetravel | Validates queries and schema |
| 0 | goose-introspection | Coordinates session discovery |
| +1 | reafference-corollary-discharge | Generates predictions |
Conservation: (-1) + (0) + (+1) = 0
Status
- Database Created: Yes
- Origin Session Marked: 20260108_22
- Sessions Tracked: 376
- Messages Indexed: 6,215
- Total Tokens: ~3.59M
Related Skills
- duckdb-timetravel - Temporal queries on session data
- duckdb-ies - Interaction entropy analysis
- reafference-corollary-discharge - Prediction/observation matching
- fswatch-duckdb - Real-time file monitoring with DuckDB
Cat# Integration
This skill maps to Cat# = Comod(P) as a bicomodule:
Forward Reference
- unified-reafference (canonical B3 Poset)