The Agent Skills Directory

[SAFE]: No security issues detected. The skill performs trace extraction and token counting as described in its documentation.
[SAFE]: The implementation includes robust mitigations against resource exhaustion (DoS). In src/harness_trace/extractor.py, the skill limits the scanning window for large tool outputs (VERIFY_SCAN_TAIL, REVIEW_SCAN_HEAD) and uses non-backtracking regex patterns to prevent quadratic processing time on blank-line-heavy output.
[SAFE]: The skill uses Pydantic models for structured data validation and standard JSON parsing, ensuring safe handling of session log content.
[SAFE]: Data ingestion is limited to local session logs and project files. The skill does not perform network operations or execute arbitrary code from the processed data.
[SAFE]: The PermissionEventData model demonstrates security awareness by explicitly requiring callers to redact secrets before populating the action field.

harness-trace