The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/safe_pack.py utility executes shell commands via subprocess.run to invoke the repomix CLI and the internal scan_secrets.py script. These executions use list-based arguments without a shell (shell=False is the default), which effectively mitigates command injection risks from user-supplied directory paths or configuration filenames.
[DATA_EXPOSURE]: The skill is specifically designed to mitigate data exposure. The scan_secrets.py script identifies potential credentials and explicitly replaces the matched content with a [REDACTED] placeholder in all outputs. This ensures that logs or reports generated by the tool do not accidentally leak the secrets they were meant to protect.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests and processes untrusted data from a target directory, the risk of indirect prompt injection is mitigated by the tool's design. The scanning logic is based on fixed regex patterns, and the output provided back to the agent is sanitized (redacted), preventing malicious instructions within scanned files from influencing the agent's execution flow.

repomix-safe-mixer