The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes logic to verify which companion skills are installed by executing an ls command on various local directory paths (e.g., ~/.claude/skills/, ~/.agent/skills/).- [EXTERNAL_DOWNLOADS]: The skill recommends the installation of missing companion skills using npx skills add from the AbsolutelySkilled vendor repository. These resources are associated with the skill's official publisher.- [PROMPT_INJECTION]: The skill has an inherent attack surface for indirect prompt injection as it is designed to ingest and process untrusted data from security incidents, such as breach reports and indicators of compromise (IOCs), which could contain malicious instructions.
Ingestion points: User-provided incident descriptions and forensic data processed within the IR framework (SKILL.md).
Boundary markers: No specific delimiters or boundary markers are utilized to separate untrusted user input from the agent's internal logic.
Capability inventory: The skill leverages several diagnostic and administrative capabilities, including shell commands, cloud CLIs (AWS CloudTrail, GCP Logging), and package managers (npm, pip) as documented in the playbooks (incident-playbooks.md).
Sanitization: There is no evidence of explicit sanitization or input validation for user-provided data before it is incorporated into the incident response workflow.

security-incident-response