The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from existing resumes (PDF/Text) and LinkedIn profiles to perform analysis and reformatting. This creates a surface for indirect prompt injection if a processed document contains hidden instructions.
Ingestion points: prompts/beautify.md (resume files) and prompts/linkedin-import.md (web-fetched or pasted profile content).
Boundary markers: The skill uses a structured YAML schema (schema/resume-data.md) to isolate and validate data before processing, which serves as a mitigation.
Capability inventory: The skill primarily performs text analysis and generates local HTML files.
Sanitization: Instructions explicitly require the agent to confirm all data with the user before final rendering.
[DYNAMIC_EXECUTION]: The skill generates HTML files that include a small JavaScript snippet to provide an 'editable mode' (using contenteditable and window.print()).
Evidence: The logic is defined in prompts/editable-version.md and injected into the final HTML output.
Risk Assessment: The generated code is static, derived from local templates, and used exclusively for the skill's primary functionality. It does not load external scripts or execute arbitrary code from untrusted sources.

resume-skill