The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided or externally sourced "options" (e.g., A, B, C) to generate narrative consequences.
Ingestion points: The primary input is the set of options provided to the agent, which may originate from untrusted sources like pull request descriptions or external data.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" blocks to prevent the agent from being influenced by instructions embedded within the options or the code/docs it is tasked to read.
Capability inventory: The agent is instructed to read local code, documentation, and interact with the /skill:prototype extension to verify behaviors.
Sanitization: There is no evidence of input validation, escaping, or filtering of the content within the options or the files being analyzed before the agent incorporates them into its reasoning context.

then-what