The Agent Skills Directory

Path Traversal Protection: The skill implements a centralized path-traversal guard in scripts/_utils.py. The safe_resolve function ensures that all file operations, such as reading templates, writing records, and updating project configurations, are strictly contained within permitted directory roots. This implementation correctly handles cross-platform path separators and resolves symlinks before performing containment checks.
Sensitive Data Detection: The skill includes a dedicated utility, scripts/scan_sensitive_content.py, which is designed to detect and flag high-confidence PII—such as email addresses, phone numbers, and national identifiers—before records are written to disk or transmitted. It also includes conditional logic to detect internal-only URLs in public repositories, reducing the risk of accidental data exposure.
Secure Template Rendering: ADR generation is performed via scripts/render_template.py, which uses regular expression-based placeholder substitution. This approach avoids the use of dynamic execution functions like eval(), significantly reducing the risk of code injection during the rendering phase.
Input Ingestion Surface: The adopt-template lifecycle ingests user-supplied markdown files to derive ADR structures. While this processing of external data represents a potential surface for indirect prompt injection, the skill addresses this risk through mandatory normalization steps and the integration of the sensitive content scanner.
Dependency Management: All external dependencies, such as pyyaml, jsonschema, and atheris, are standard, versioned packages from established registries. These are appropriate for the skill's stated purpose of document validation and fuzz testing.

adr-author