The Agent Skills Directory

[SAFE]: The skill defines secure engineering practices for telemetry, emphasizing structural integrity and data privacy.
[CREDENTIALS_UNSAFE]: The skill actively prevents credential exposure by providing a clipError function and SECRET_PATTERNS regular expressions designed to redact Bearer tokens, Slack tokens, and OpenAI API keys from logs and traces.
[DATA_EXFILTRATION]: Instructions explicitly warn against logging personally identifiable information (PII) and mandate bounding field lengths to prevent accidental leakage of sensitive or unbounded subprocess output.
[INDIRECT_PROMPT_INJECTION]: The skill mitigates potential injection from untrusted data (such as third-party error messages) by requiring pure, unit-tested sanitizers and schema-level validation before any external string is recorded.
[EXTERNAL_DOWNLOADS]: Telemetry export is configuration-gated and relies on industry-standard OpenTelemetry protocols, ensuring diagnostic data is only transmitted to authorized, user-defined endpoints.

observability