Menü
Consolidate and clean up existing brain memory. Run this periodically (like a sleep cycle) to dedupe, decay, merge, and relocate memories to vault.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
Remove old pi session files to reclaim disk space
Testing-focused coding workflow. Extends code-assist with enhanced test generation, coverage analysis, and test quality gates.
Systematic debugging workflow for diagnosing and fixing issues. Follows a structured approach to isolate problems, form hypotheses, and verify fixes.
Generate documentation from code or scratch - READMEs, API docs, changelogs, and user guides. Complements codebase-summary which analyzes existing code.
Extract durable learnings and preferences from the current conversation for memory capture. Use this after a conversation to capture what should be remembered.
Track skill usage with correctness scoring, generate reports, and identify unused or misused skills for pruning.
| name | cleanup-memory |
| description | Consolidate and clean up existing brain memory. Run this periodically (like a sleep cycle) to dedupe, decay, merge, and relocate memories to vault. |
| kind | sop |
Run a "brain sleep cycle" on your existing memories:
When to use:
extract-memoryDifference from extract-memory:
extract-memory = "What should I remember from this conversation?" (captures new)cleanup-memory = "Clean up my existing memories" (organizes old)~/.rho/brain/brain.jsonl)true)last_consolidation) — last_consolidation | <ISO timestamp> | <duration>1)~/.pi/agent/sessions/)10)high)memory_consolidate.last_consolidated_at)List memory by type and count totals:
Determine the lower bound timestamp for session mining.
Extract durable learnings/preferences from user messages in matched sessions.
Confidence policy:
mediumRun brain action=decay to age out old entries.
Identify duplicates/superseded/stale entries and merge candidates.
Importance levels:
Apply the 30-day test: "Would this change a decision I make 30 days from now?"
Auto-remove categories:
Move reference-heavy entries to vault (useful knowledge that doesn't need injection every turn):
Set/update checkpoint timestamp after successful completion.
Report counts before/after, mining window, added/skipped entries, decayed/removed/merged/relocated counts, vault notes created.
Run periodically (suggested: daily or weekly):
/skill run cleanup-memory
With custom parameters:
/skill run cleanup-memory since=7d max_new_entries=5