The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to analyze and 'internalize' external repositories, which involves reading untrusted content such as README files, source code, and git logs. Ingestion points: Repository files and git metadata ingested during analysis phases (Phases 1, 3). Boundary markers: None; the agent is instructed to process all content to extract mindsets. Capability inventory: Reading and writing files, listing directories, and executing shell commands. Sanitization: None. A maliciously crafted repository could include instructions to poison the generated research documents or influence agent behavior.- [COMMAND_EXECUTION]: The skill instructs the agent to run various shell commands like git log, ls, and find to analyze target repositories. Additionally, it provides a Python script scripts/scaffold-notes.py and encourages users to execute it. While the script uses standard libraries for file system operations, executing commands and scripts based on or alongside external data presents an inherent risk surface.

repo-deep-dive