The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided inputs, such as technical receipts and project evidence, which represents an ingestion surface for untrusted data. However, the skill explicitly includes instructions to treat these inputs as private and incorporates a review step to ensure that private identifiers, metrics, and backlinks do not appear in the final output. Given that no dangerous tools or system-level capabilities are invoked, the risk is negligible.
[NO_CODE]: The skill consists entirely of instructional markdown. It does not contain any scripts, shell commands, or external package dependencies. All logic is contained within the natural language instructions and a local reference file.

blog