The Agent Skills Directory

[PERSISTENCE]: The skill provides instructions to append environment variables to the user's shell profile (~/.zshrc) to ensure API keys persist across shell sessions (SKILL.md).
[COMMAND_EXECUTION]: The pipeline relies heavily on executing various security-oriented CLI tools including subfinder, httpx, nuclei, and ffuf, piping their output through shell utilities like grep, jq, and awk (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill fetches data from several external security research services including crt.sh, ProjectDiscovery's Chaos API, and GitHub's API (SKILL.md).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from arbitrary target domains through a complex chain of tools.
Ingestion points: Data enters through curl requests to external discovery services and via crawlers like katana, waybackurls, and gau (SKILL.md).
Boundary markers: No explicit boundary markers or instructions are provided to tools to ignore potentially malicious embedded content within the discovered subdomains or URLs.
Capability inventory: Processed data is piped into operations including nuclei vulnerability scanning, python3 script execution (SecretFinder), and directory fuzzing with ffuf (SKILL.md).
Sanitization: Sanitization is limited to basic output filtering via jq, sed, and grep, which does not prevent attacks against more complex tool parsers.
[CREDENTIALS_UNSAFE]: The setup process encourages users to store sensitive API keys for third-party services (VirusTotal, SecurityTrails, Censys, Shodan) in local configuration files (~/.config/subfinder/config.yaml) and shell environment variables (SKILL.md).

web2-recon