[ronpandolfi]

Audit and fix pass over all 11 agent skills against the current source (gpCAM master, fvgp 4.11.1). Each claim about the API was checked against gpcam/gp_optimizer_base.py or the fvgp 4.11.1 source before the text was changed. There's one commit per skill.

Corrections

• experiment-designer: the text said tell() replaces data by default. The default is append=True (gp_optimizer_base.py, def tell).
• prior-mean-functions: the description said the default prior is a "flat zero prior". The body (correctly) says it's mean(y_data).
• acquisition-functions: reconciled with references/choosing-an-acquisition-function.md (they had drifted apart), fixed the multi-task EI incumbent, and documented ask(args=...).
• gp2scale-advanced: Client() snippets now sit under if __name__ == "__main__":.
• kernel-designer: the deep-kernel example trains with method="adam", which is what its own notes recommend.
• uncertainty-calibration: posterior_mean returns a dict, and the in-loop retrain uses n_restarts.

Consistency with 8.7.0

• Version headers are bumped to 8.7.x / fvgp 4.11.x across the board (six skills still said 8.4.x).
• experiment-designer now covers the multi-task modes of the single GPOptimizer. The noise, prior-mean and kernel skills note the cartesian-mode shapes.
• The hyperparameter-index convention and the 3- vs 4-argument args dispatch table are now in one canonical place (experiment-designer). The other skills point to it instead of repeating it, which cuts about 2.5k words (−10%).

Interaction with #57: #57 moves skills/ to gpcam/skills/. Git's rename detection should merge these edits cleanly in either order. After the move, the relative link ../../examples/GPOptimizer_LogAndLogit.ipynb in transformed-optimizers-advanced needs one more ../.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.60%. Comparing base (a21385a) to head (e3840a0).

Additional details and impacted files

@@           Coverage Diff           @@
##           master      #58   +/-   ##
=======================================
  Coverage   98.60%   98.60%           
=======================================
  Files           8        8           
  Lines        1509     1509           
=======================================
  Hits         1488     1488           
  Misses         21       21           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.