The Agent Skills Directory

[SAFE]: The skill is a formatting utility designed to create self-contained HTML documents with strong privacy protections.
It mandates that all CSS and JS be inlined, strictly prohibiting the use of CDNs, external fonts, or any network requests in the generated output files.
This ensures that artifacts are offline-capable and effectively prevents data exfiltration via external asset requests.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions provide clear guidance on handling sensitive information and preventing unauthorized data exposure.
Sensitive material is explicitly directed to be stored in private or uncommitted directory locations to avoid accidental publication via automated repository tools.
The routing instructions favor local delivery to the user using secure rendering tools rather than external hosting services.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided data to generate reports, creating a theoretical ingestion surface.
Ingestion points: Untrusted user content processed in SKILL.md for transformation into HTML briefs or reports.
Boundary markers: The instructions focus on visual output invariants; while explicit delimiters for input are not described, the output is static.
Capability inventory: File writing and the SendUserFile tool for local rendering.
Sanitization: The skill emphasizes privacy and local rendering, mitigating the risk of downstream exploitation as the output triggers no automated network or system actions.

pretty-html