The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation recommends the use of the --dangerously-skip-permissions flag. This flag disables critical human-in-the-loop safety checks, allowing the AI agent to execute potentially destructive shell commands or system modifications without user confirmation.
[PERSISTENCE_MECHANISMS]: The skill utilizes Claude Code's hooks feature to automatically execute shell commands or scripts during lifecycle events (e.g., Stop, PostToolUse). This creates a mechanism for persistent, automated command execution whenever the CLI tool is active.
[EXTERNAL_DOWNLOADS]: The skill references external dependencies and scripts, such as a secret_scanner.py located in a hardcoded local directory (C:/Users/renat/skills/cred-omega/scripts/) and custom Node.js MCP servers, which are not provided in the skill package and constitute unverified external code.
[INDIRECT_PROMPT_INJECTION]: The skill presents a large attack surface for indirect prompt injection due to its core function of processing project data with high-privilege tools. Ingestion points: Reads CLAUDE.md, project source files, and git output. Boundary markers: The instructions lack explicit directives to the agent to ignore or delimit instructions found within the project data. Capability inventory: Full Bash shell access, filesystem Read and Write capabilities, and the ability to spawn sub-agents. Sanitization: No validation or escaping of external content is described.
[METADATA_POISONING]: The risk: none claim in the metadata is deceptive given the encouragement to bypass security permissions and the implementation of automated persistence hooks.

claude-code-expert