The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The SKILL.md file contains instructions to execute code directly from a remote GitHub repository by piping it to the Python interpreter (`curl -sL ... | python3
search `). This practice allows for the execution of arbitrary remote code that bypasses local file inspection and versioning controls.
[EXTERNAL_DOWNLOADS]: The scripts/re_index.py file automatically downloads a JSON data file (entries.json) from a remote GitHub URL during the first execution or when the local cache expires. While the source corresponds to the author's repository, it establishes a dependency on remote data.
[INDIRECT_PROMPT_INJECTION]: The skill implements an info command that fetches and displays the content of README files from external, untrusted repositories. This provides a vector for indirect prompt injection, as malicious instructions within those READMEs could be interpreted by the AI agent.
Ingestion points: External repository README files fetched via the GitHub API in scripts/re_index.py (github_readme function).
Boundary markers: The output uses simple text delimiters ("--- README ---"), but lacks explicit instructions to the agent to ignore any embedded commands within the fetched content.
Capability inventory: The skill can perform shell command execution (git clone), manage local cache files, and perform authenticated network requests via urllib.
Sanitization: There is no evidence of sanitization or filtering applied to the fetched README content before it is output to the agent's context.
[COMMAND_EXECUTION]: The Python script uses the subprocess module to execute system commands, specifically git clone for downloading repositories and gh auth token to retrieve authentication credentials from the GitHub CLI. These calls interact with the host system's shell environment and tools.

find-reverse-engineered