The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses tools like WebFetch and WebSearch to ingest data from external security blogs, hacker forums, and vulnerability databases. Since the agent also possesses Write and Edit capabilities for local files, this configuration presents an indirect prompt injection surface where malicious instructions embedded in researched web content could potentially influence the agent's behavior or file operations.
Ingestion points: WebFetch and WebSearch tools are used to gather data from external research topics.
Boundary markers: The instructions provide structure for output but do not include explicit delimitations or warnings to ignore instructions embedded within the fetched research data.
Capability inventory: The skill has access to Read, Edit, Write, Grep, Glob, WebFetch, and WebSearch tools (defined in SKILL.md).
Sanitization: No specific sanitization or validation of the content retrieved from external sources is described.
[EXTERNAL_DOWNLOADS]: The skill instructions and reference files point to a wide variety of external security-related domains for research. These include vulnerability databases (cve.mitre.org, nvd.nist.gov, exploit-db.com), government agencies (cisa.gov, fbi.gov, nsa.gov), and major security vendors (mandiant.com, crowdstrike.com, securelist.com, microsoft.com, talosintelligence.com). The skill also references security journalism sites and community archives (phrack.org, 2600.com). These references are consistent with the skill's stated purpose of cybersecurity research.

researchers-security