The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): Several setup scripts automate the installation of third-party packages from PyPI.
scripts/setup-agentops.sh: Executes pip install -U agentops.
scripts/setup-phoenix.sh: Executes pip install openinference-instrumentation-google-adk google-adk arize-phoenix-otel.
scripts/setup-weave.sh: Executes pip install opentelemetry-sdk opentelemetry-exporter-otlp-proto-http.
DATA_EXFILTRATION (MEDIUM): The primary purpose of this skill is to log agent interactions, including user inputs and LLM responses, to external services (BigQuery, AgentOps, Arize Phoenix, Weights & Biases).
Ingestion points: templates/bigquery-analytics-config.py and other config files ingest LLM_REQUEST and LLM_RESPONSE data.
Capability inventory: Exports data via network operations to third-party endpoints (e.g., https://trace.wandb.ai, app.agentops.ai).
Boundary markers: Absent in templates.
Sanitization: The skill provides a commendable example of a sanitize_content function in templates/bigquery-analytics-config.py to redact PII and keys, but its implementation is optional and depends on user configuration.
COMMAND_EXECUTION (LOW): Scripts like scripts/setup-bigquery-analytics.sh and scripts/setup-cloud-trace.sh execute system commands (gcloud, bq, gsutil) to configure cloud infrastructure and modify IAM permissions (e.g., granting roles/bigquery.dataEditor). while legitimate for setup, these require high-privilege credentials.

observability-patterns