The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill's installation instructions for Linux environments utilize sudo to perform administrative actions such as modifying system keyrings and installing software via the package manager.
Evidence: sudo gpg --no-default-keyring --keyring /usr/share/keyrings/k6-archive-keyring.gpg --keyserver hkp://keyserver.ubuntu.com:80 --recv-keys C5AD17C747E3415A3642D57D77C6C491D6AC1D69 in SKILL.md.
Evidence: sudo apt-get update && sudo apt-get install k6 in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of security keys and software from external repositories during the setup process.
Evidence: Fetches GPG keys from keyserver.ubuntu.com and software from the official k6 repository at dl.k6.io.
Evidence: Directs users to download binaries from the official Vegeta GitHub repository releases page.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external files such as JavaScript test scripts and target configuration files, which presents a surface for indirect prompt injection attacks.
Ingestion points: Reads external files including script.js, targets.txt, and body.json using the read tool.
Boundary markers: No boundary markers or specific instructions are provided to help the agent distinguish between its instructions and content inside the processed scripts.
Capability inventory: The skill allows the agent to use bash, read, and write tools, which can be leveraged to execute code or manipulate the filesystem based on instructions embedded in the scripts.
Sanitization: There is no evidence of validation or sanitization of the scripts or target files before they are passed to execution tools like k6 or vegeta.

load-testing