Trace one analysis
SkillDev toolsTrace an analytical claim to its receipt, source, query, result, and run evidence. Use when the user asks where a number came from, wants to verify a query, or needs to inspect provenance.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Trace one analysis skill
What this skill tells your AI
The instructions your AI receives, as published by ai-analyst-lab/ai-analyst in .claude/skills/trace-analysis/SKILL.md and read by ahel’s review.
Resolve the claim through these records:
- The output states the claim.
- The receipt identifies the source, data snapshot, query, and artifacts.
- The trace records how the run unfolded.
- Logs record specific system and query events.
- The query and result provide execution evidence.
Use helpers.evals.trace.inspect_receipt to identify missing fields and initial query risks. Then inspect the analytical details that code cannot infer from syntax alone: population, filters, time window, timezone, exclusions, grain, join cardinality, missingness, referential integrity, freshness, reconciliation, and whether the claim stays within the evidence.
For a query with a join and aggregate, test the grain directly. Compare the candidate total with a defensible reference query when one exists. A query that runs successfully may still answer the wrong question.
Report missing evidence as missing. A complete receipt makes an analysis inspectable. It does not make the analysis correct.
Signals
- GitHub stars
- 297
- Forks
- 137
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
trace-analysis- Source
- github.com/ai-analyst-lab/ai-analyst