The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external identifiers (domains, URLs, IPs) from potentially hostile sources. This creates a surface for indirect prompt injection where malicious data could attempt to manipulate the agent's behavior during investigation.
Ingestion points: External target strings and URLs provided to the /cti, /cti-recall, and /cti-pivot commands.
Boundary markers: Absent. The instructions do not define specific delimiters or instructions to ignore embedded commands within the analyzed data.
Capability inventory: The skill can orchestrate sub-agents via the --deep flag, execute shell commands (T2 markers), utilize MCP tools (T1 markers), and write case information to the local intel_engine/cases/ directory.
Sanitization: Absent. There is no mention of sanitizing or validating external identifiers before they are incorporated into tool inputs or sub-agent prompts.
[COMMAND_EXECUTION]: The skill documentation describes the use of a registration script (scripts/register.sh) to install tools. Additionally, it features a --yolo flag specifically intended to suppress user confirmation prompts, allowing the agent to execute actions autonomously without manual oversight.
[PERSISTENCE]: The skill automates the installation of persistent commands by symlinking files from the skill's commands/ directory into the user's ~/.claude/commands/ directory, ensuring they remain available across sessions.

dym-cti-expert-investigate-in-claude-code