The Agent Skills Directory

[NO_CODE]: The skill consists entirely of Markdown documentation and protocol definitions; no executable code or scripts are included in the skill package.
[EXTERNAL_DOWNLOADS]: Automated scanners identified https://url3.com as malicious. This URL is used as a placeholder in documentation templates in SKILL.md and references/message-templates.md. While the usage is contextual, its presence in a research-oriented skill could pose a risk if an agent literalizes the templates.
[PROMPT_INJECTION]: The defined research pipeline is vulnerable to indirect prompt injection based on the provided playbook. 1. Ingestion points: The skill instructions direct specialists (ExaAI and Firecrawl) to process external web data. 2. Boundary markers: The communication protocol uses Markdown headers but lacks explicit delimiters or 'ignore' instructions for ingested content. 3. Capability inventory: The agents utilize CLI tools for web search, crawling, and local file system writes for findings and reports. 4. Sanitization: There are no protocols defined for sanitizing or validating untrusted data before it is relayed to other agents or the orchestrator.

agentic-research