The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external transcript data. It lacks specific delimiters or 'ignore' instructions for the LLM context, creating a potential surface for indirect prompt injection if a transcript contains malicious instructions. However, the risk is mitigated by explicit instructions to 'Confirm consent, participant labeling, and redaction needs' and to separate verbatim evidence from interpretation.
[COMMAND_EXECUTION]: The skill references a local CLI tool memi for research tasks. This tool is part of the 'Memoire' ecosystem developed by the author, and its usage is strictly aligned with the skill's primary research purpose.
[DYNAMIC_EXECUTION]: The documentation describes a codegen pipeline that transforms processed interview data into React components. This is a legitimate functional feature of the research toolchain and does not involve executing untrusted code in a sensitive context.

interview-research