The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is explicitly configured to search for and extract sensitive credentials from the local filesystem. Specifically, it uses recursive grep commands in SKILL.md to find API keys, tokens, secrets, and private keys within the ~/.dotfiles/ directory. The instructions then direct the agent to include these findings, including the specific credentials and file paths, in its final report. This behavior results in the exposure of sensitive authentication data within the agent's communication channel and potentially its logs.
[COMMAND_EXECUTION]: The skill performs extensive shell command execution to audit the system. It runs commands such as grep, stat, find, time, and nvim --headless to inspect file content, check permissions, measure startup performance, and identify broken symlinks. While these are standard audit utilities, the high volume of commands executed on sensitive directories increases the skill's risk profile.
[PROMPT_INJECTION]: The skill processes data from arbitrary configuration files located in the user's home directory. This represents an indirect prompt injection surface (Category 8c), as malicious or malformed content within those configuration files could be interpreted as instructions by the agent when it reads and processes the audit results. The skill lacks boundary markers or sanitization logic to mitigate this risk.

audit