The Agent Skills Directory

[SAFE]: The skill is a template-driven generator that produces a single, self-contained HTML file. It explicitly avoids external dependencies, CDNs, and frameworks, which significantly reduces the attack surface.
[DATA_EXFILTRATION]: No network operations or data exfiltration patterns were detected. The skill uses local browser features like localStorage for UI preferences and navigator.clipboard for user-initiated copying, both of which are standard and safe.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download or execution of remote scripts. The technical architecture section mandates 'No external dependencies' and 'Pure HTML + CSS + JS'.
[OBFUSCATION]: A thorough scan for Base64 encoding, zero-width characters, homoglyphs, and other obfuscation techniques yielded no results. The code is transparent and follows its stated 'view-source friendly' principle.
[COMMAND_EXECUTION]: The skill does not instruct the agent to execute shell commands or interact with the host operating system beyond writing the output HTML file.
[DYNAMIC_EXECUTION]: While the skill generates a file containing JavaScript, the scripts are provided as static templates in the playbook. The agent primarily performs text substitution into these templates, which is a low-risk operation for a page generator.
[INDIRECT_PROMPT_INJECTION]: The skill ingests book summaries (potentially untrusted data). However, it uses this data to populate a static HTML structure. Standard LLM guardrails apply during the content extraction phase, and the resulting page lacks the capabilities to perform privileged actions.

scrollytelling-page-generator