The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands and scripts (e.g., git ls-files, grep, bash scripts/scan-secrets.sh) to perform security audits. These commands are scoped to the local repository and do not interact with external networks.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill accesses sensitive paths (like .env files or files containing secret), it does so strictly for local auditing purposes. There are no network operations detected that would send this data to a remote server.
[REMOTE_CODE_EXECUTION]: All scripts are contained within the skill package, and no external code is downloaded or executed during the workflow.
[PROMPT_INJECTION]: The instructions do not contain patterns attempting to override agent behavior or safety guidelines. The !command syntax is not used maliciously; the skill provides legitimate developer-oriented git workflows.

safe-repo