Skill Bundles

Read the source. Install what you trust.

Each skill bundle packages a reusable agent behavior — a prompt, supporting files, and evaluation criteria. Browse the public catalog, review the full source, then install a private copy you can edit and experiment with.

Published bundles
108
Total installs
0
Average quality
70/100

Browse bundles

108 published bundles ready to inspect and install

Skill bundlev1.0.0

Transfer Eval Design

Build evals that test whether RL training on task A improved performance on related task B

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Risk Tier Classification

Classify agent skills by risk level (read-only vs. write vs. financial vs. external-facing) and apply appropriate controls

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Audit Trail For RL Decisions

Log every decision an RL agent makes in production with sufficient context for post-hoc review

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Deployment Gating Pipeline

Build eval-gated deployment pipelines where RL-trained models must pass benchmarks before production

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Data Exfiltration Prevention

Monitor and prevent agents from leaking sensitive data through tool calls

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Skill Security Audit

Static and dynamic analysis of agent skill code for security vulnerabilities

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Deceptive Alignment Detection

Test whether agents behave differently when they believe they're being evaluated vs. not

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

RL Alignment Auditing

Verify that the policy optimizes for the intended objective, not a proxy

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Action Space Sandboxing

Restrict agent actions to prevent irreversible or harmful operations

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Safe Exploration Constraints

Define and enforce hard constraints on what agents can do during training rollouts

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Reward Hacking Red Teaming

Systematically find ways an agent could game the reward function

0 installs
70/100 quality
Compatibility not listed
Inspect bundle
Skill bundlev1.0.0

Rollback And Versioning

Maintain and switch between agent versions when new RL training degrades performance

0 installs
70/100 quality
Compatibility not listed
Inspect bundle