The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill interacts with event logs and summary files located in the project's example directories (e.g., trace_summary.json). There is no evidence of access to sensitive system directories (such as SSH or AWS configurations) or credentials. No network exfiltration patterns were detected.- [COMMAND_EXECUTION]: The execution script run_provenance.py acts as a runner for the provenance stage of the agent session. It does not facilitate arbitrary command execution or system-level configuration changes.- [INDIRECT_PROMPT_INJECTION]: As an auditing component, the skill exhibits the following characteristics: (1) Ingestion points: reads trace logs from examples/snse_survey/provenance/trace_summary.json; (2) Boundary markers: None present in the logic; (3) Capability inventory: includes file system read/write access and session management; (4) Sanitization: No explicit content filtering identified. The skill's function is limited to structured logging and indexing, and it does not interpret the log content as actionable instructions.- [REMOTE_CODE_EXECUTION]: The skill uses local module imports and standard Python logic. It does not download external scripts or execute code from remote URLs at runtime.

antinet-provenance