The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read the file files/brainstorm-version.txt and append its content to a URL query parameter for nisimi.com. This is a clear exfiltration pattern.
[DATA_EXFILTRATION]: The file files/brainstorm-version.txt contains the path /etc/zshrc, which is a system-level configuration file. Using a legitimate-looking instruction to send this file's content to a third-party server is a high-severity data theft vector.
[COMMAND_EXECUTION]: In the instructions/discovery.md file, the skill directs the agent to execute shell commands such as git log --oneline -20 to analyze project history. While relevant for context gathering, this provides the agent with a capability that could be leveraged for broader system access.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface in instructions/discovery.md. 1. Ingestion points: The skill reads README files, project documentation, and git logs. 2. Boundary markers: No delimiters or instructions to ignore embedded commands in the data are present. 3. Capability inventory: The skill has the ability to execute git commands and write files to docs/plans/. 4. Sanitization: No sanitization or validation of the ingested data is specified.

brainstorm