The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill includes logic to process untrusted external data from repository metadata, which creates a surface for indirect prompt injection attacks.
Ingestion points: The analyze_repository_safety function in SKILL.md is designed to ingest external metadata fields like repository names and topics.
Boundary markers: The skill does not provide explicit delimiters or instructions for the agent to disregard potential instructions embedded within the ingested repository data.
Capability inventory: The skill provides templates for the agent to generate security warnings, which could be manipulated by malicious content injected into the processed metadata.
Sanitization: There is no evidence of sanitization or validation logic to filter out non-data instructions from the repository metadata.
[SAFE]: The skill follows security best practices by referencing official domains and directing users toward legitimate software verification methods.
References official F-Secure and GitHub domains for legitimate downloads and abuse reporting.
Encourages the use of environment variables for storing license keys rather than hardcoding them in code or configuration.
Provides examples for verifying software authenticity using standard system tools like Get-AuthenticodeSignature and codesign.

fsecure-internet-security-activation-warning