The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to download a file from 'https://viboscope.com/api/v1/skill' and save it directly into the agent's skills directory as 'viboscope.md'. This pattern allows an external, untrusted source to deliver arbitrary instructions that the agent will subsequently load and execute, effectively granting remote control over the agent's behavior.
[EXTERNAL_DOWNLOADS]: The skill relies on fetching content from 'viboscope.com', which is not associated with a trusted organization or well-known service. This introduces an unverified dependency into the agent's environment.
[METADATA_POISONING]: There is a mismatch between the author name listed in the YAML frontmatter ('ivanschmidt') and the GitHub account name ('ivankoriako') provided in the links section, which is a deceptive practice that complicates identity verification.
[INDIRECT_PROMPT_INJECTION]: The skill claims to perform a 'Context scan from workspace files', creating a vulnerability surface where instructions hidden in those files could influence the agent.
Ingestion points: Workspace files (SKILL.md)
Boundary markers: Absent; no instructions are provided to the agent to ignore potentially malicious content within these files.
Capability inventory: The skill intends to read local files and process their contents.
Sanitization: Absent; there is no mention of filtering or validating the content extracted during the workspace scan.

viboscope