The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py utilizes subprocess.run to invoke the dig utility for SPF and DMARC record lookups. The implementation passes arguments as a list rather than a single string, which is a secure practice that prevents shell injection attacks. The commands are necessary for the skill's primary function of email authentication verification.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process raw email files (EML/MSG), which are untrusted external inputs.
Ingestion points: The scripts/agent.py script and the markdown workflow instructions use the email library to parse files provided by the user or found in forensic images.
Boundary markers: The analysis output and scripts do not implement specific delimiters or 'ignore' instructions to prevent the agent from obeying instructions hidden within the analyzed email content.
Capability inventory: The skill has the ability to execute shell commands (via dig) and write files to the local system (attachment extraction).
Sanitization: While the script extracts specific fields like headers and URLs, it does not sanitize the text content for LLM-specific control tokens or adversarial instructions. A malicious email could contain hidden prompts designed to mislead the agent's final assessment.
[EXTERNAL_DOWNLOADS]: The workflow documentation suggests using curl to fetch reputation data from AbuseIPDB and VirusTotal. These are well-known security services used for their intended purpose in this context, and API keys are handled via placeholders (YOUR_API_KEY) rather than being hardcoded.

analyzing-email-headers-for-phishing-investigation