The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted codebases, creating a vulnerability to indirect prompt injection.
Ingestion points: Code files read via tldr search, tldr extract, and tldr context.
Boundary markers: The instructions lack markers to distinguish code data from agent instructions.
Capability inventory: The skill uses Bash to execute commands, providing a significant capability surface.
Sanitization: No sanitization is specified for the code content returned by the tool.
[EXTERNAL_DOWNLOADS]: The tldr semantic index command downloads embedding models (e.g., bge-large-en-v1.5) from external repositories to enable semantic search capabilities.
[COMMAND_EXECUTION]: The skill leverages the Bash tool to execute various tldr CLI commands for navigating and querying the codebase.

tldr-code