The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is highly susceptible to Indirect Prompt Injection (Category 8) due to its core function of processing untrusted external data.
Ingestion points: exiftool (reference/image.md) processes metadata from potentially malicious images; dig and whois (reference/domain.md) ingest data from attacker-controlled DNS records and WHOIS databases; curl (reference/domain.md) fetches data from the Wayback Machine API.
Boundary markers: Absent. The skill provides no instructions to the agent to treat external tool outputs as untrusted or to ignore embedded instructions within that data.
Capability inventory: The skill allows Bash and Write (SKILL.md), enabling arbitrary command execution and file system modification if the agent is manipulated by processed data.
Sanitization: Absent. There are no patterns for escaping or validating the content retrieved from external OSINT sources before it is processed by the agent or passed to other shell commands.
External Dependencies (MEDIUM): The skill relies on several third-party tools that are not standard system utilities, which may require external downloads or installations.
Evidence: References to sherlock (reference/social.md), subfinder and amass (reference/domain.md), and exiftool (reference/image.md).
Command Execution (LOW): The skill heavily utilizes shell commands for its operations. While intended for OSINT, the use of unzip on untrusted Office documents (reference/social.md) and strings on arbitrary files carries inherent risks if the underlying utilities have vulnerabilities.

osint