The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by design. It dynamically selects and loads modules into the agent's context based on untrusted external signals.
Ingestion points: The selection logic in modules/selection-strategies.md uses user_input and detected_files as signals to trigger module loading.
Boundary markers: No explicit instructions or delimiters are provided to the agent to ignore instructions found within the dynamically loaded modules.
Capability inventory: The patterns provided allow for the modification of the agent's instructions at runtime through load_module and unload_module calls.
Sanitization: No sanitization of the input signals (keywords or file paths) is demonstrated in the provided code patterns.
[NO_CODE]: The skill consists entirely of Markdown files and does not include any executable scripts or binaries. All code provided exists as snippets within the documentation for reference purposes.

progressive-loading