From soundcheck
Detects training pipelines that ingest external data without integrity gating. Use when auditing dataset ingestion, fine-tuning scripts, or web-scraped data curation.
How this skill is triggered — by the user, by Claude, or both
Slash command
/soundcheck:training-data-poisoningThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Protects against malicious or low-quality examples being introduced into training or
Protects against malicious or low-quality examples being introduced into training or fine-tuning datasets. Poisoned data can embed backdoors, degrade accuracy, or skew model behavior in ways that are difficult to detect after training completes.
Flag the vulnerable code, explain the risk, and suggest a fix establishing these properties. Translate to the data-loading and validation libraries of the audited file — use that stack's documented hashing, schema, and dataframe APIs; do not import a recipe from a different stack.
Confirm the response:
npx claudepluginhub thejefflarson/soundcheck --plugin soundcheckValidates, deduplicates, and tracks provenance of training data to detect poisoning, bias, and privacy violations before model training.
Detects compromised or backdoored models from unverified sources, floating tags, or unreviewed registries. Use when downloading pre-trained models, loading from registries, integrating third-party LLM providers, or managing automated model updates.
Detects AI/ML security vulnerabilities like unsafe model deserialization in PyTorch/Joblib/NumPy, prompt injection in LLM prompts, and risks in Jupyter notebooks or ML pipelines.