The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the agent to use sudo to read restricted system files, including Nginx/Apache access logs and systemd unit files. It also suggests using sudo to access shell histories of other users, including the root user.- [DATA_EXFILTRATION]: The instructions involve searching for and accessing highly sensitive files that often contain secrets, such as .env files, process environment blocks (/proc/<pid>/environ), and shell history files (.bash_history).- [COMMAND_EXECUTION]: The skill provides numerous shell command templates for the agent to run locally, including grep, rg, find, and curl.- [INDIRECT_PROMPT_INJECTION]: The skill establishes a large attack surface by requiring the agent to process and analyze untrusted content from the local system, such as web server logs and application source code.
Ingestion points: Reads host logs, source code files, and system configurations.
Boundary markers: Includes instructions to only reference token prefixes and database IDs in findings to avoid leaking full credentials.
Capability inventory: Uses shell commands for file searching and inspection, along with database client tools for schema introspection.
Sanitization: The skill advises against including raw token material in reports and suggests redaction patterns for logging.

aegis-api-token-security