The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external sources, such as X (Twitter) and various GitHub repositories, which presents a surface for indirect prompt injection.
Ingestion points: Processes tweets via the xc client and external repository content found through gh search to derive skill instructions in SKILL.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are implemented when processing external content.
Capability inventory: The skill allows writing to the file system, performing GitHub operations, and executing repository scripts via pnpm.
Sanitization: External content is not sanitized or escaped before being used to author instructions.
[COMMAND_EXECUTION]: Uses local shell commands for catalog management, including pnpm for manifest building and testing, and grep for searching local files. The skill also utilizes a pre-authenticated X client (xc) to retrieve social media data.
[EXTERNAL_DOWNLOADS]: Fetches content from GitHub and X (Twitter) for research and catalog entry creation. These operations target established services and do not involve untrusted third-party servers.

create-paperclip-bundled-skill