The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts scripts/github_index.py and scripts/pk_index.py fetch repository source data as tarballs from the GitHub API (api.github.com). This is standard functionality for a repository indexing tool using a well-known service.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it processes untrusted content from external repositories and includes it in a generated index file.
Ingestion points: Repository file content is read and processed in scripts/github_index.py and scripts/pk_index.py to extract headings, YAML metadata, and code symbols.
Boundary markers: None; extracted content is placed directly into markdown tables or lists without delimiters or warnings for the agent to ignore embedded instructions.
Capability inventory: The skill can perform network requests to the GitHub API and write files to the local file system.
Sanitization: No security-focused filtering is performed on the extracted strings, allowing potentially malicious instructions in a repository to enter the agent's context.
[REMOTE_CODE_EXECUTION]: Automated scans flagged a pattern of piping curl output to python3. Analysis confirms this is a benign command template provided in the documentation for decoding Base64 file content from the GitHub API. The Python code is a static one-liner that performs only data transformation and does not execute the content of the downloaded files.

building-github-index