The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/github-codebase-search.py uses bunx @morphllm/morphmcp@latest to download and execute code from an external provider at runtime. This introduces a risk of executing unverified or malicious code from a third-party source.
[CREDENTIALS_UNSAFE]: The script passes the MORPH_API_KEY directly as a command-line argument using the --env flag. On many operating systems, environment variables and command-line arguments are visible to all users via process listing tools like ps.
[COMMAND_EXECUTION]: The tool invokes external binaries and scripts (mcporter, bunx) using the subprocess.run function, interpolating user-supplied search queries and repository information into the command string.
[PROMPT_INJECTION]: The documentation explicitly instructs the agent not to read the source code of the scripts (DO NOT read script source code), which prevents the agent from verifying the tool's behavior or identifying potential security issues.
[PROMPT_INJECTION]: The skill retrieves and processes untrusted data from public GitHub repositories, creating a surface for indirect prompt injection.
Ingestion points: GitHub repository content fetched via the MorphLLM API in scripts/github-codebase-search.py.
Boundary markers: The skill lacks explicit instructions to the model to ignore any instructions found within the retrieved code snippets.
Capability inventory: The script has the ability to execute shell commands and download packages via subprocess.run.
Sanitization: There is no evidence of sanitization or safety checks on the content retrieved from external repositories.

github-codebase-search