The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (check-freeze.sh and gstack-paths) and standard shell commands (mkdir, echo, git, cd, pwd). These are used for administrative tasks, path resolution, and maintaining a local analytics log of skill usage within the user's home directory.
[DYNAMIC_CONTEXT_INJECTION]: The skill documentation uses shell script blocks that the agent is expected to execute to set up the environment. These commands are transparent, relate directly to the skill's stated purpose of locking a directory, and do not incorporate unvalidated remote data.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill maintains a local usage log at ~/.gstack/analytics/skill-usage.jsonl. This data remains on the local filesystem and is not transmitted externally. The script explicitly suppresses errors during this logging process to avoid interfering with the user experience.
[INDIRECT_PROMPT_INJECTION]: The skill ingests a directory path from the user via AskUserQuestion. The input is used to resolve a path on the filesystem. While the skill uses eval on a local binary (gstack-paths), it is a local component of the toolchain and the user-provided path is handled via standard shell resolution (cd/pwd) rather than being directly evaluated as code.

freeze