The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources, specifically during competitor analysis (scraping websites and reviews via Firecrawl as mentioned in references/competitor.md) and community management (processing member introductions and feedback as mentioned in references/community.md). There are no explicit instructions for the agent to sanitize this input or use boundary markers to prevent the execution of instructions embedded in the external data. The agent has powerful capabilities including bash and run_command, which could be exploited if malicious instructions are successfully injected via these external data sources.
Ingestion points: External competitor websites, customer reviews, and community platform messages (Discord, Slack, Skool).
Boundary markers: Absent in the provided instructions; the agent is not instructed to use delimiters or ignore embedded instructions when processing external content.
Capability inventory: File system access (view_file, write_to_file, replace_file_content) and shell execution (bash, run_command).
Sanitization: No explicit sanitization, escaping, or validation logic is defined for the external content before it is processed by the agent.

growth