[Chore] Faster notebooks & tdhook skill

[Xmaster6y]

What does this PR do?

Key insights about the PR.

Linked Issues

• Closes [Bug] Some notebooks are not executed #42
• Closes [Feature] Create agent skill for using the library #35

Checklist

• I have read the CONTRIBUTING guide.
• I have added tests for my changes if needed.
• I have updated the documentation if needed.

---

Summary by cubic

Speed up docs builds by never executing notebooks and keeping pre-run outputs. Add a tdhook agent skill with reference docs, and surface Bilinear Probing and Skills in the README.

• New Features

  • Added tdhook skill docs (SKILL.md) with references: API, tutorials, issues, releases, file structure.
  • README: new Skills section and Bilinear Probing link.

• Docs/Build

  • Set nbsphinx_execute to "never" to avoid running notebooks during builds.
  • Removed nb-clean pre-commit hook to preserve notebook outputs.
  • Updated notebooks with execution counts and sample outputs across methods and tutorials (IG, steering, linear/bilinear probing, dimension estimation, TorchRL PPO, chess).

^{Written for commit 0b64a3a. Summary will update on new commits.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.19%. Comparing base (e37469c) to head (0b64a3a).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #43   +/-   ##
=======================================
  Coverage   97.19%   97.19%           
=======================================
  Files          39       39           
  Lines        2426     2426           
=======================================
  Hits         2358     2358           
  Misses         68       68           

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

[cubic-dev-ai]

5 issues found across 17 files

Confidence score: 4/5

• This PR looks safe to merge; issues are limited to documentation snippets in skills/tdhook/references/tutorials.md with low severity.
• Most significant impact is example code that won’t run as written (e.g., undefined DEV and missing torch.nn/tensordict imports), which can confuse readers but won’t affect runtime behavior.
• Pay close attention to skills/tdhook/references/tutorials.md - fix missing imports and define DEV in the setup snippet.

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="skills/tdhook/references/tutorials.md">

<violation number="1" location="skills/tdhook/references/tutorials.md:72">
P3: The Steering Vector example uses TensorDict without importing it, so the snippet fails when copied. Add the missing tensordict import.</violation>

<violation number="2" location="skills/tdhook/references/tutorials.md:177">
P3: The Chess Board Saliency example uses TensorDict without importing it. Add the tensordict import to keep the snippet runnable.</violation>

<violation number="3" location="skills/tdhook/references/tutorials.md:201">
P3: The Activation Patching example uses TensorDict without importing it. Include the tensordict import in the snippet.</violation>

<violation number="4" location="skills/tdhook/references/tutorials.md:241">
P3: The Inline Adapters example uses nn.Module without importing torch.nn. Add the missing torch.nn import so the snippet runs.</violation>

<violation number="5" location="skills/tdhook/references/tutorials.md:284">
P3: The setup snippet references DEV without defining it, which breaks the example. Provide a default value for DEV before using it.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}