The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes a shell script (scripts/probe_asset_credentials.sh) that accesses and sources sensitive shell configuration files, including ~/.bashrc, ~/.zshrc, ~/.bash_profile, and ~/.zprofile. Access to these files is categorized as sensitive file exposure.\n- [DATA_EXFILTRATION]: Instructions in SKILL.md mandate that the agent execute a credential probe and include the output (revealing the presence or absence of environment variables like TRIPO_API_KEY, GEMINI_API_KEY, and ELEVENLABS_API_KEY) in its final report, exposing environmental metadata to the agent's output.\n- [COMMAND_EXECUTION]: The skill is designed to orchestrate the execution of multiple local scripts, including Bash shell scripts, Python scripts, and Node.js utilities (e.g., audit_reference_report.py and probe_asset_credentials.sh) across the skill and its required siblings.\n- [PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection by actively searching for and loading instructions from multiple external SKILL.md and reference files across various filesystem paths (e.g., ~/.claude/skills/, ~/.agents/skills/).\n
Ingestion points: Sibling SKILL.md files and reference markdown files loaded from various relative and absolute system paths.\n
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are used when loading these external files.\n
Capability inventory: The skill can execute shell scripts, Python code, and Node.js scripts.\n
Sanitization: No sanitization or validation of the content of the loaded files is performed before they are integrated into the agent's context.

threejs-game-director