The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The skill automatically loads and applies instructions, resources, and configurations from ~/.claude/skills/CORE/USER/SKILLCUSTOMIZATIONS/OSINT/ to override default behavior. This dynamic loading from a specific file path allows for potential persistent instruction injection if the directory is accessible to other processes.
[Indirect Prompt Injection] (LOW): The skill's primary function is to ingest untrusted data from the web, which creates a significant attack surface for indirect prompt injection.
Ingestion points: Data is retrieved from dozens of external sources including crt.sh, social media, DNS records, and various threat intelligence APIs (VirusTotal, AlienVault OTX, etc.).
Boundary markers: Absent. The workflows do not specify delimiters or instructions to ignore embedded commands in the retrieved data.
Capability inventory: The skill has the ability to spawn sub-agents (Task DSL), write files to the local system (scratch and history directories), and perform network operations via curl.
Sanitization: Absent. The methodology focuses on synthesis and reporting without explicit sanitization of external content.
[Command Execution] (LOW): SKILL.md executes a hardcoded curl command to localhost:8888 for notifications. While directed at a local endpoint, it demonstrates active use of subprocess spawning.
[External Downloads] (LOW): The skill facilitates and recommends the use of numerous external tools and websites for data collection. While these are industry-standard OSINT tools, the agent acts as an automated bridge to these external services.

OSINT