TAAC Platform API Inspection

SkillDocs & knowledge

Inspect TAAC/Taiji training checkpoints and scalar metrics through an authenticated platform session.

Available today. Use it from your connected AI after setup.

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