The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script scripts/validate_architecture.py to perform static analysis on architecture documents. This script uses standard Python libraries (re, json, argparse, pathlib) and does not perform network operations or execute arbitrary strings.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied architecture documents via the validation script.
Ingestion points: The scripts/validate_architecture.py script reads the content of a user-provided Markdown file.
Boundary markers: The skill's instructions explicitly warn the agent in references/01-workflow-and-decision-gates.md and references/11-ai-ml-and-agentic-systems.md to treat untrusted text reaching models with tools as hostile instructions and to separate data from authority.
Capability inventory: The skill uses file-read capabilities and local command execution via Python. It does not have network-write capabilities in its scripts.
Sanitization: The validation script performs structural and regex-based checks rather than executing the contents of the document.
[REMOTE_CODE_EXECUTION]: The skill references several external technical standards and documentation repositories (e.g., NIST, OWASP, Google SRE, C4 model). These references are informative and originate from well-known, trusted organizations.
[DATA_EXPOSURE]: The templates and examples provided (e.g., assets/adr-template.md, examples/worked-example-order-platform.md) use standard placeholders and fictional data for demonstration. No hardcoded credentials or sensitive environmental paths were found.

system-architecture-harness