The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): The file contains purely instructional documentation for shell script workflows. There are no attempts to override agent behavior, bypass safety filters, or extract system prompts.
[DATA_EXFILTRATION] (SAFE): No sensitive file paths, credential patterns, or network communication commands are present. The examples use standard temporary file creation (mktemp) and local directory operations.
[REMOTE_CODE_EXECUTION] (SAFE): There are no commands that download or execute remote scripts (e.g., curl|bash). While it mentions tools like shellcheck and shfmt, it does not attempt to install them or execute them on untrusted data.
[COMMAND_EXECUTION] (SAFE): The skill provides templates for shell scripts but does not execute them. The recommended practices, such as using set -euo pipefail and proper variable quoting, actually mitigate common shell script vulnerabilities.
[INDIRECT_PROMPT_INJECTION] (SAFE): While the skill defines how to process data, it explicitly mandates input validation and quoting, which are primary defenses against injection attacks in shell environments.

shell-workflow