The Agent Skills Directory

[COMMAND_EXECUTION]: The script_tester.py utility uses subprocess.run to execute Python scripts found in the target skill's directory. This is used for runtime validation of help commands and sample data processing. The implementation follows best practices by using argument lists instead of shell strings and includes execution timeouts to prevent resource exhaustion.
[DYNAMIC_EXECUTION]: The skill implements a framework that dynamically discovers and executes Python code from a user-provided directory path. It mitigates potential risks by performing a static analysis pass using the ast module to verify syntax and ensure that the code only imports standard library modules before execution.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes content from external skill packages.
Ingestion points: Reads files such as SKILL.md (containing YAML frontmatter), README.md, and Python scripts from a target directory.
Boundary markers: The tool does not use specific delimiters to isolate untrusted content from its instructions, though it does treat the content as data for validation.
Capability inventory: The skill has the ability to execute code via subprocess.run and perform file system operations.
Sanitization: It employs yaml.safe_load for frontmatter parsing and ast.parse for script validation, providing significant protection against malicious content during the analysis phase.
[METADATA_POISONING]: The documentation and script headers claim the skill has 'Zero External Dependencies' and uses the 'Python standard library only'. However, skill_validator.py and quality_scorer.py both depend on the external PyYAML library for parsing frontmatter. While PyYAML is a common and reputable package, the claim is technically inaccurate.

skill-tester