The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill recommends a "Token Top10" workflow where the agent processes user interview transcripts to extract insights. This creates a surface for indirect prompt injection.
Ingestion points: SKILL.md suggests feeding interview transcripts into AI for analysis.
Boundary markers: There are no instructions provided to delimit transcripts or ignore potentially malicious content within them.
Capability inventory: The skill is documentation-only and does not include scripts; however, it assumes the agent has the capability to interact with external tools like Linear or other issue trackers.
Sanitization: No sanitization or validation methods for the transcript data are mentioned.
[NO_CODE]: The skill consists entirely of Markdown files, including guides, templates, and metadata. It does not contain any scripts, executables, or configuration files that would trigger automated code execution or system changes.

gingiris-user-interview