The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The toolkit demonstrates a vulnerability surface for indirect injection via data ingestion. Findings evidence: 1. Ingestion points: The importData function and vulnerability tracking functions in SKILL.md ingest external JSON and finding details. 2. Boundary markers: Absent; there are no instructions to the agent to treat ingested data as untrusted. 3. Capability inventory: Use of document.write, innerHTML, and insertAdjacentHTML. 4. Sanitization: Absent; user-controlled data is interpolated directly into HTML templates.
[DYNAMIC_EXECUTION]: The toolkit employs dynamic execution patterns by generating HTML content at runtime. The generateCustomReport function assembles a report using string concatenation and uses document.write to execute it in a new window. Other functions like displayVulnerabilitiesPaginated and addCustomOWASPTest use innerHTML and insertAdjacentHTML to inject content into the active document.
[EXTERNAL_DOWNLOADS]: The toolkit instructions guide users to download code from an external source: https://github.com/Cyber-Autopsie/PenTest-Toolkit-Pro.git.

pentest-toolkit-pro-html