The Agent Skills Directory

External Downloads & Command Execution (MEDIUM): The skill instructs the agent to download files from arbitrary URLs found during web scraping. It provides templates for curl and Python requests execution.
Evidence: Step 6 explicitly provides shell commands (curl -o "{save_path}" "{logo_url}") and Python code blocks for file operations.
Risk: If {save_path} or {logo_url} are not sanitized, an attacker could achieve command injection (e.g., using ; or & in the string). Additionally, the skill lacks validation of the downloaded content's integrity or type beyond basic file extension checks.
Indirect Prompt Injection (LOW): The skill's core functionality involves ingesting data from untrusted external sources which could contain malicious instructions.
Ingestion points: Website HTML, GitHub README files, and social media metadata (Step 1, 2, 3, 4).
Boundary markers: Absent; the skill does not suggest using delimiters or instructions to ignore embedded commands in the source data.
Capability inventory: File system write access (via curl or Python), network requests, and browser automation via MCP Playwright.
Sanitization: Absent; the skill does not provide mechanisms to sanitize the content of processed files or the metadata extracted from the web.
Data Exposure (LOW): The skill allows writing to a user-defined save_path. Without path validation, this could be used for path traversal to overwrite sensitive system files or configuration files (e.g., .bashrc or .ssh/authorized_keys).

extract-project-logo