The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from external sources including URLs, GitHub repositories, and media transcripts.
Ingestion points: External URLs for repositories, articles, media files, and skill packages provided by the user.
Boundary markers: The instructions do not specify any delimiters or safety markers to isolate the intaken content from the agent's internal reasoning or instructions.
Capability inventory: The skill uses git clone to download code, writes analysis reports to the local file system, and calls the yt CLI for transcript processing. It also loads local project documentation such as AGENTS.md and NORTH_STAR.md into context.
Sanitization: There is no evidence of sanitization or filtering of the intaken content before it is processed by the agent.
[REMOTE_CODE_EXECUTION]: The skill facilitates the acquisition of external code via git clone and mentions a lifecycle for installing "skill packages" using a skill-builder tool. Downloading and installing code from untrusted external repositories presents a risk of executing malicious payloads.
[COMMAND_EXECUTION]: The skill utilizes shell commands for environment resolution and external tool interaction, including git rev-parse --show-toplevel and the yt CLI for media processing.
[DATA_EXPOSURE]: The skill reads local configuration files from the user's home directory (~/.config/lev/config.yaml) and the project root (.lev/config.yaml) to resolve workshop paths and settings.

lev-intake