The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web and uses it to update local files and GitHub issues, creating a surface for indirect instructions to influence agent behavior.\n- Ingestion points: The skill executes five web searches to gather data on AI agent frameworks and security (W02).\n- Boundary markers: The instructions lack explicit delimiters or warnings to treat external search results as untrusted data.\n- Capability inventory: The agent has the ability to write to local wiki and log files (llmwiki/html/overstack-vs-world-30d.html, llmwiki/wiki/log.md, llmwiki/wiki/sources/ISSUES.md) and execute GitHub CLI commands (gh issue create, gh issue comment) to modify project tracking.\n- Sanitization: There is no evidence of content filtering or validation for the data retrieved from web searches before it is interpolated into the reports or used to create tracker issues.

frontier-scan