Detect and block outbound data exfiltration in agent tool calls.
npx skills add https://github.com/brucebanner010198-commits/DevSecOps-Agency --skill dlp
Please help me install this Agent Skill. Skill: dlp Source: https://github.com/brucebanner010198-commits/DevSecOps-Agency/tree/main/skills/dlp Command: npx skills add https://github.com/brucebanner010198-commits/DevSecOps-Agency --skill dlp
Outbound data leaks from agent tool calls pose a risk to credentials, secrets, and sensitive information. This Skill provides automated detection and protection to prevent data exfiltration.
Configure the DLP system to scan outbound tool calls for sensitive data and block or redact matches in real time.
High-intent search queries and answers about installing and using this skill.
Detecting split-secret leakage requires a chain-correlation window that tracks outbound data across multiple agent calls, correlating payloads to identify fragmented secrets distributed across separate requests before enforcing redaction.
You can balance security with legitimate workflows by configuring per-project waivers for sanctioned data flows, applying redaction tokens instead of hard blocking, and integrating with existing secrets vaults and runtime hooks.
NLP-based scanning detects PII and PHI in outbound agent payloads by analyzing arguments, URL paths, headers, and query strings for sensitive language patterns, applying static matching alongside natural language processing to identify exposed personal data.
Outbound DLP scanning supports a configurable custom term library per project, allowing teams to define project-specific sensitive terms and patterns for detection across all outbound agent tool calls.
The best way to detect split-secret leakage is using a chain-correlation window that monitors outbound tool calls over time, correlating fragmented data across multiple requests to catch secrets intentionally split across separate payloads.
Use per-project waivers for outbound data loss prevention when you have sanctioned data flows requiring legitimate transmission of sensitive information, allowing approved calls to bypass blocking while maintaining scanning on all other traffic.
ThiNepoImplement human-in-the-loop approval workflows for LangChain agents.
gpt-cmdrRepair HEC-RAS geometry issues with RasFixit and validate via RasCheck.
matteocervelliScan codebases for secrets, vulnerable dependencies, and insecure patterns.
gke-labsAudit Kubernetes network egress, RBAC, and volume mounts for data exfiltration risks.
