| name | retrospectives |
| description | The reflection and iterative improvement layer, managing project and session retrospectives. |
Retrospectives Skill
The Retrospectives skill provides a structured way to capture, organize, and act upon project and session reflections. It transforms raw feedback into actionable ledger tasks to drive iterative improvement.
Core Capabilities
The system organizes retrospectives into a time-indexed hierarchy (Year $\rightarrow$ Month $\rightarrow$ Day) to ensure historical continuity and easy retrieval.
Retrospective Capture
Supports three levels of reflection:
- Task-based: Post-mortems for specific high-impact tasks.
- Daily: Session-end reflections on daily progress.
- Weekly: Higher-level architectural and process reviews.
The Reflection Schema
Every retrospective captures a structured state:
- Well: What went right? (Successes to replicate)
- Not Well: What failed? (Pain points to resolve)
- Start: New behaviors or tools to adopt.
- Stop: Counter-productive patterns to cease.
- Continue: Stable practices to maintain.
- Improvements: Direct actionable goals.
Actionable Ledger
The skill integrates with the project ledger. When a retrospective identifies a critical failure or improvement, it can spawn a ledgerTask, linking the theoretical reflection to an actual work item.
Commands
save_retro
Persists a structured retrospective to the sovereign memory.
- Arguments:
date (YYYY-MM-DD), type (task/day/week), id, content (RetroSchema).
get_retros_for_period
Retrieves all retrospectives within a specific date window.
- Arguments:
start_date, end_date.
create_ledger_task
Transforms a retrospective finding into a formal ledger task.
- Arguments:
description, priority, source_retro_id.
Implementation Details
- Architecture: Two-tier Python implementation (FastMCP $\rightarrow$ RetrospectivesLogic).
- Storage: Filesystem-based hierarchy using JSON documents.
- Organization: Time-series layout for optimized retrieval and historical auditing.