The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes grep, find, sed, and sort through a shell script (find.sh) to perform file and text searches. These are standard development tools and the script uses set -euo pipefail for robust error handling. No arbitrary command injection via unvalidated arguments was detected as the script primarily constructs patterns for these specific binaries.
[DATA_EXFILTRATION]: The skill implements a robust EXCLUDE_DIRS list including node_modules, .git, .env, and venv to prevent the agent from accidentally reading or exposing sensitive configuration, credentials, or version control metadata during broad searches.
[PROMPT_INJECTION]: The SKILL.md contains strict operational instructions ('Never scan the repo with AI', 'Results cap at 50 lines') that act as internal guardrails. These instructions are designed to improve tool reliability and prevent hallucination rather than bypassing safety protocols.
[REMOTE_CODE_EXECUTION]: No external downloads or remote script executions were detected. All operations are local to the repository using the provided find.sh script.

find-code