The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/engine.py file uses subprocess.run to invoke the ast-grep (sg) tool. This is used for structural analysis of the code. The parameters passed to the tool are derived from internal rule definitions and validated against a set of known IDs, which prevents arbitrary command injection.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes user-provided Bash scripts.
Ingestion points: User-provided Bash scripts are read into the agent's context for analysis and refactoring (as described in SKILL.md).
Boundary markers: Present. The skill employs a structured workflow using markdown lists for reporting and manual verification steps.
Capability inventory: The skill can read/write files and execute subprocesses (via the scripts/bash-shorten.py script).
Sanitization: The automated rewriter uses structural parsing (ast-grep), which validates Bash syntax but does not filter natural language instructions embedded within comments or strings.
[EXTERNAL_DOWNLOADS]: The documentation references external open-source tools including ast-grep, sd, ripgrep, and fd. These are well-known developer utilities hosted on public GitHub repositories and are used for legitimate code processing tasks.

bash-shortening