The Agent Skills Directory

[SAFE]: The skill performs accessibility audits using bundled and local scripts (scripts/contrast.py, fitness-config.py). These scripts use standard Python libraries to parse CSS and calculate contrast ratios without performing network requests or executing arbitrary code.
[SAFE]: All external references target the author's own domain (jeffbailey.us), which is consistent with the skill's stated purpose and author context. No suspicious external dependencies or remote script downloads were detected.
[SAFE]: The skill implements strict confidence thresholds (confidence >= 7/10) and requires specific file:line evidence for all findings, which reduces the risk of hallucinations or misleading reports.
[INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface because it is designed to ingest and analyze untrusted frontend code (HTML, CSS, JS) provided by the user.
Ingestion points: HTML templates, CSS files, and JavaScript components in the repository under review.
Boundary markers: The skill does not explicitly define boundary markers for untrusted code, but instructs the agent to focus on specific code patterns for audit purposes.
Capability inventory: The skill uses python3 to execute local scripts for analysis and write_file to generate a markdown report (docs/accessibility-review.md).
Sanitization: No specific sanitization of code content is performed beyond regex-based parsing in the contrast script. However, because this ingestion is the primary purpose of the tool and the resulting actions are limited to local analysis and reporting, the risk is considered safe.

review-accessibility