// This skill should be used when the user asks to "learn from past sessions", "extract lessons", "find what worked", "identify patterns in my workflow", "what mistakes did I make", "what decisions were made", or wants to synthesize wisdom and actionable insights from conversation history.
Use when auditing, extracting, pruning, or writing Claude Code memory files. Covers memory file format, MEMORY.md index conventions, staleness scoring, claim extraction heuristics, destination routing, mutation rules, and archive conventions.
This skill should be used when the user asks to "check git history", "trace code changes", "find when something was added", "correlate commits with sessions", "find hotspot files", or needs to mine git repository history for insights. Only applicable when the current project is a git repository.
This skill should be used when the user asks to "analyze conversation history", "parse JSONL files", "read past sessions", "search conversation logs", "find what happened in a session", or needs to work with Claude Code's .jsonl conversation format. It provides the canonical parsing infrastructure for echo-sleuth agents.
| name | experience-synthesis |
| description | This skill should be used when the user asks to "learn from past sessions", "extract lessons", "find what worked", "identify patterns in my workflow", "what mistakes did I make", "what decisions were made", or wants to synthesize wisdom and actionable insights from conversation history. |
| version | 0.1.0 |
When analyzing conversations, extract insights in these categories (ordered by durability — values first):
Comparative preferences and priority orderings — "X is better than Y".
Signal sources:
What to capture: The value choice, both sides of the comparison, and the reasoning. Values are the most durable type of memory — they survive codebase rewrites.
What was chosen, why, and what alternatives were rejected.
Signal sources:
AskUserQuestion tool calls + the user's response in the next tool_resultplanContent field on user records, ExitPlanMode tool calls)What to capture: The decision, the rationale, the alternatives considered, and the context (what problem it solved).
What went wrong, root cause, and how it was fixed.
Signal sources:
is_error: trueWhat to capture: What failed, why it failed, what fixed it, how to avoid it next time.
Approaches that worked well and could be reused.
Signal sources:
pr-link records)What to capture: The approach, when it applies, why it worked.
Approaches that failed, were abandoned, or caused problems.
Signal sources:
What to capture: What was tried, why it failed, what worked instead.
The user's preferred tools, styles, and workflows.
Signal sources:
What to capture: The preference, evidence across sessions, strength of pattern.
System design decisions, component relationships, tech stack details.
Signal sources:
What to capture: Component, its role, relationships, key decisions about it.
Issues that keep coming back across sessions.
Signal sources:
What to capture: The problem, frequency, root cause pattern, whether it has a permanent fix.
Token usage trends, session efficiency, cost optimization opportunities.
Signal sources:
session-stats.sh output across sessions: token totals, compaction countssystem records with subtype: turn_durationWhat to capture: Trends, outliers, optimization opportunities.
When extracting insights, follow this process:
list-sessions.sh to identify relevant sessions, then extract-messages.sh and extract-tools.sh for detailPresent insights in a structured format:
## [Category]: [Brief Title]
**Sessions:** [session IDs or dates]
**Context:** [What was happening]
**Insight:** [The key learning]
**Evidence:** [Specific data points]
**Action:** [What to do differently / what to keep doing]
Good insights are:
Skip insights that are: