Event Hooks

proagent-ml-ai

2 events · 4 hooks

npx claudepluginhub diegouis/provectus-marketplace --plugin proagent-ml-ai

Risk Indicators

Executes bash commandsModifies files

Source

Defined in hooks/hooks.json

Configuration

PreToolUse

Matcher: Bash

Hooks:

command

which jq >/dev/null 2>&1 || exit 0; jq -r '.tool_input.command // empty' | { read cmd; if echo "$cmd" | grep -qiE 'rm\s+-rf\s+models/|rm\s+-rf\s+mlruns/|rm\s+-rf\s+data/'; then echo '[proagent-ml-ai] Destructive operation on ML artifacts detected. Verify this is intentional.' >&2; exit 2; fi; }

10ms

command

which jq >/dev/null 2>&1 || exit 0; jq -r '.tool_input.command // empty' | { read cmd; if echo "$cmd" | grep -qiE 'python.*train|mlflow run|python.*-m.*train'; then echo '[proagent-ml-ai] Training command detected. Ensure random seeds are set, experiment tracking is configured, and data splits prevent leakage.'; fi; }

10ms

PostToolUse

Matcher: Bash

Hooks:

command

which jq >/dev/null 2>&1 || exit 0; jq -r '.tool_input.command // empty' | { read cmd; if echo "$cmd" | grep -qiE 'python.*train|mlflow run|python.*evaluate'; then echo '[proagent-ml-ai] Training/evaluation completed. Verify experiment metrics are logged and model artifacts are saved.'; fi; }

10ms

Matcher: Write

Hooks:

command

which jq >/dev/null 2>&1 || exit 0; jq -r '.tool_input.file_path // empty' | { read fp; if echo "$fp" | grep -qiE '\.(py|ipynb)$'; then echo '[proagent-ml-ai] ML code written. Verify random seeds, data leakage prevention, and experiment tracking are included.'; fi; }

10ms

Summary

Blocks unsafe git ops like main commits, unverified pushes, blanket adds, branch deletes, and root/main edits via PreToolUse bash hooks on Bash/Edit tools. Warns risks, reminds CI, watches deploys post-push.