Agent skill
ralph-analytics
Show analytics and metrics from agent loop execution history
Install this agent skill to your Project
npx add-skill https://github.com/jmagly/aiwg/tree/main/agentic/code/addons/agent-loop/skills/ralph-analytics
SKILL.md
Al Analytics Command
Display aggregate analytics and metrics from agent loop execution history.
Instructions
When invoked, analyze agent loop data and present metrics:
-
Scan Loop History
- Load all loop records from
.aiwg/ralph/ - Load reflections from
.aiwg/ralph/reflections/ - Load debug memory from
.aiwg/ralph/debug-memory/
- Load all loop records from
-
Calculate Metrics
- Success rate: % of loops that completed successfully
- Average iterations: Mean iterations to completion
- Reflection reuse rate: % of reflections applied in subsequent loops
- Stuck loop rate: % of loops that hit stuck detection
- Escalation rate: % requiring human intervention
-
Pattern Analysis
- Most common failure types
- Most effective fix patterns
- Average time per iteration
- Quality trajectory per loop
-
Display Dashboard
- Summary metrics table
- Trend indicators (improving/stable/degrading)
- Recommendations for improvement
Arguments
--since [date]- Analyze loops from date (default: all)--loop [id]- Analyze specific loop--export [path]- Export analytics to file--brief- Show summary only
References
- @$AIWG_ROOT/agentic/code/addons/ralph/schemas/reflection-memory.json - Reflection schema
- @$AIWG_ROOT/agentic/code/addons/ralph/schemas/debug-memory.yaml - Debug memory schema
- @$AIWG_ROOT/agentic/code/addons/ralph/docs/reflection-memory-guide.md - Guide
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
research-document
Generate summaries and literature notes from research papers
research-archive
Package research artifacts for long-term archival
research-cite
Format citations and generate bibliographies
induct-research
Induct research sources into a research repository. Point at an issue, a single file, a directory of papers, or a URI and the skill reads, annotates, and files structured induction tasks — one per source. Similar to address-issues but for research corpora instead of code backlogs.
research-provenance
Query provenance chains and artifact relationships
research-quality
Assess source quality using GRADE methodology
Didn't find tool you were looking for?