The Agent Skills Directory

[PROMPT_INJECTION]: The skill demonstrates a vulnerability to Indirect Prompt Injection (Category 8). It is designed to ingest and process external, untrusted data from social media platforms which could contain malicious instructions hidden by attackers.
Ingestion points: Phase 1 (Trend Monitoring, Competitor Analysis) and Phase 3 (Community Management and Comment engagement) involve reading data from external sources.
Boundary markers: The instructions do not define boundary markers or explicit warnings for the agent to ignore instructions embedded within the analyzed content.
Capability inventory: The skill possesses significant capabilities via the bash, git, and filesystem tools (File ops) listed in the Copilot CLI Operations section.
Sanitization: There is no evidence of sanitization, filtering, or validation of the external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill explicitly requests access to powerful system tools including bash and git in the "Herramientas disponibles" section. While these tools are common in developer-focused agent environments, they provide a broad capability for system interaction that increases the impact of a potential prompt injection.

tiktok-strategist