The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content, such as patent claims and evidence documents, to populate analysis tables. While this creates a potential surface for indirect prompt injection, the risk is mitigated by the skill's specific instructions to perform verbatim citation without paraphrasing, which limits the model's likelihood of interpreting the data as instructions.
[SAFE]: The skill implements a security best practice for data export. It explicitly instructs the agent to escape potential formula injection characters (=, +, -, @) with a single quote before writing values to CSV or spreadsheet formats.
[SAFE]: No network activity or external data transmission was detected. All operations are confined to the local filesystem and the agent's internal reasoning.

claim-chart