The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system commands to validate the structure and syntax of other Agent Skills.\n
Evidence: In scripts/validate.py, subprocess.run is used to invoke ruby for YAML parsing, bash -n for shell script linting, and python3 -m py_compile for Python syntax checks.\n
Evidence: In scripts/test_validate.py, subprocess.run is used to execute regression tests against the validator script.\n- [DYNAMIC_EXECUTION]: The validator script uses an embedded Ruby script to perform strict structural analysis.\n
Evidence: scripts/validate.py contains a RUBY_YAML string constant containing Ruby code that is executed via ruby -e to check for duplicate YAML keys.\n- [INDIRECT_PROMPT_INJECTION]: As an auditing tool, the skill is designed to ingest and process untrusted instructional prose from external Agent Skill files, which could theoretically contain malicious instructions targeting the auditor.\n
Ingestion points: The Audit action reads the content of SKILL.md and all referenced resources in the target directory (e.g., references/, scripts/, assets/).\n
Boundary markers: The skill instructions emphasize following a strict quality rubric and reporting contract, but does not implement specific technical boundary markers for the raw text it analyzes.\n
Capability inventory: The skill has the ability to read and execute sub-processes for validation purposes using the subprocess module.\n
Sanitization: Performs structural validation and syntax checking on input files, which acts as a form of format-level sanitization.

x9-skill-creator