The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes a PowerShell script (scripts/update_references.ps1) to download and update DOTA2 API references by cloning a remote repository from GitHub (https://github.com/BigCiba/vscode-dota2-tools.git). This source is a third-party repository and is not associated with the trusted vendor list.\n- [PRIVILEGE_ESCALATION]: The instructions in SKILL.md for updating references suggest executing the PowerShell script with the -ExecutionPolicy Bypass flag. This command-line argument bypasses the local script execution policy, allowing the script to run regardless of the system's security configuration.\n- [COMMAND_EXECUTION]: The skill promotes a workflow that requires the execution of multiple system commands and tools, including git for repository management, python for running internal search and detection scripts, and project-specific tools like npm and npx tstl for building addon assets.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection as it reads and interprets untrusted files from the user's project environment.\n
Ingestion points: Reads project configuration files (package.json, tsconfig.json) and source code files (Lua, TypeScript, XML, KV) during detection and development tasks.\n
Boundary markers: While the skill defines architectural boundaries for the developer, it lacks technical delimiters or explicit instructions to the model to ignore potential commands embedded within the data it processes.\n
Capability inventory: The skill environment allows for broad file system access (read/write), subprocess execution (Python, Node.js tools), and network requests (Git).\n
Sanitization: No specific sanitization or content validation is performed on the data ingested from the project files before they are presented to the agent context.

dota2-custom-game-dev