The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it ingests and processes untrusted content from the Nexudus platform, such as blog comments, community messages, and help desk tickets.
Ingestion points: Untrusted data enters the agent context through CLI list and get commands across almost all entity reference files (e.g., blogpostcomments.md, communitymessages.md, helpdeskmessages.md).
Boundary markers: The skill mandates the use of the --agent flag, which returns data in a structured JSON envelope, providing clear boundaries between data and instructions.
Capability inventory: The skill allows the agent to create, update, and delete entities on the Nexudus platform using the nexudus CLI (documented in SKILL.md and all files in references/entities/).
Sanitization: The CLI automatically redacts PII using deterministic tokens (e.g., «PII:EMAIL:a3f2b1c9»), which the agent is instructed to pass back without modification.
[NO_CODE]: The provided skill files consist entirely of Markdown documentation and instructions. No executable scripts, binaries, or source code were included in the analyzed package.
[SAFE]: The skill uses established security practices for AI agents, such as prohibiting the agent from requesting credentials and relying on manual user login (nexudus login). All external resources and APIs referenced belong to the official Nexudus platform.

nexudus