The Agent Skills Directory

[PERSISTENCE]: The skill instructs the user to add environment variable exports to shell configuration files such as .bashrc or .zshrc. This is a persistence mechanism used to maintain platform configuration across multiple terminal sessions.
[EXTERNAL_DOWNLOADS]: The instructions direct the user to download a third-party Java extension from an unverified community GitHub repository (BuffaloWill/burpsuite-project-file-parser). This introduces a supply chain risk by requiring the installation of unvetted software from a non-trusted source.
[CREDENTIALS_UNSAFE]: The skill is designed to search and extract data from Burp Suite project files, which inherently contain sensitive information such as authentication headers, session cookies, and API keys. Reading these files exposes credentials and secrets to the agent's processing context.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests data from .burp project files, which contain arbitrary HTTP request and response traffic originating from external and potentially untrusted web servers.
Boundary markers: The instructions recommend using head -c 50000 and jq for truncation, but these are manual guidelines for the agent and do not constitute a secure boundary against adversarial content.
Capability inventory: The agent utilizes the Bash and Read tools to execute commands and process the captured traffic data.
Sanitization: The skill lacks automated sanitization or filtering of the retrieved HTTP traffic, creating a surface where malicious instructions embedded in captured responses could be processed by the agent.

burpsuite-project-parser