The Agent Skills Directory

[SAFE]: The skill demonstrates a high degree of security awareness by explicitly instructing the agent never to include passwords or tokens in generated links. It correctly identifies that WebFig session tokens are stored in localStorage and warns against including them in diagnostics or shared files.
[SAFE]: The URL fragment catalog in references/webfig-url-scheme.md proactively identifies and labels 'doit' or 'action' containers which could trigger state changes on the router if opened. The skill instructs the agent to avoid these for human review links, focusing only on verified read-only views.
[EXTERNAL_DOWNLOADS]: The validation documentation references quickchr and centrs, which are tools provided by the same author (tikoci) to facilitate the creation of isolated lab environments for testing. These are documented as vendor-controlled resources for reproduction and verification purposes.
[INDIRECT_PROMPT_INJECTION]: While the skill contains a large catalog of technical identifiers (URL fragments) that the agent processes, the risk of injection is mitigated by the clear categorization of these strings and the explicit instructions on how to safely build and verify links. The skill does not ingest untrusted runtime data into a high-privilege execution environment.

routeros-webfig