TAAC Platform API Inspection
SkillDocs & knowledgeInspect TAAC/Taiji training checkpoints and scalar metrics through an authenticated platform session.
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 TAAC Platform API Inspection skill
What this skill tells your AI
The instructions your AI receives, as published by puiching-memory/taac_2026 in .agents/skills/taac-platform-api-inspection/SKILL.md and read by ahel’s review.
Use structured API data for exact checkpoint and metric values. Read API access and response layout when fetching live data. For an already downloaded payload, use the response-layout section directly.
Return data relevant to the request: checkpoint inventory for checkpoint questions; first/latest values, extrema with steps, and relevant intervals for training analysis. Summarize large payloads near the data; if a tool already wrote a complete local file, inspect it rather than repeating the request.
Interpret metrics using the experiment and run that produced them. For current Symbiosis diagnostics, consult the experiment page. Validate suspected overfitting against held-out AUC/LogLoss, and match a metric's best step to an existing checkpoint before recommending a checkpoint. Missing metrics are not zero; fixed percentage thresholds do not fit every scalar.
Use the existing authenticated session without exposing cookies, tokens, or authorization headers. If login is required, ask the user to log in; continue analysis of any data already available. Publishing, deleting, cancelling, or submitting platform jobs requires the user's request for that action. Keep downloaded platform payloads and screenshots out of commits.
Signals
- GitHub stars
- 315
- Forks
- 63
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
taac-platform-api-inspection- Source
- github.com/puiching-memory/taac_2026