The Agent Skills Directory

[SAFE]: The skill includes a validation script (scripts/check_standard_markdown.py) that detects active Hugo residue and local absolute file paths in generated Markdown, helping to prevent unintended data disclosure.\n- [SAFE]: The inventory_hugo_rules.py script implements path validation (safe_user_path) to ensure output files are written only within the current workspace, mitigating path traversal risks.\n- [SAFE]: Instructions in SKILL.md explicitly restrict the agent's capabilities by forbidding network operations, package installations, and the execution of Hugo templates or Go template expressions.\n- [PROMPT_INJECTION]: The skill processes untrusted Hugo content (Ingestion points: content/ and data/ directories in the site root). It implements boundary markers by instructing the agent to replace unresolved shortcodes with 'Conversion note' blocks. The capability inventory includes file system reads and execution of the skill's own Python scripts. Sanitization is performed via the check_standard_markdown.py script to detect leftover active syntax.

hugo-to-markdown