The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data, including user-uploaded PDFs, resumes, and screenshots, to extract information for resume generation. This ingestion point creates a surface for indirect prompt injection if malicious instructions are embedded in the processed documents.
Ingestion points: User messages, attachments, and project files containing experience data (specified in SKILL.md).
Boundary markers: The skill does not explicitly define delimiters to separate untrusted data from its system instructions, though it mentions a claim-evidence ledger for fact-checking.
Capability inventory: The skill has the capability to generate HTML files and read local template assets from the file system.
Sanitization: The instructions include rules for handling missing fields (using placeholders) and resolving conflicts between data sources, but do not specify escaping or sanitization of the content before inclusion in HTML templates.
[COMMAND_EXECUTION]: The file references/page-balance-qa.md contains a JavaScript snippet designed to be executed in a browser's developer console. This script measures DOM elements to ensure A4 page alignment and content density. While the script is technically benign and intended for layout quality assurance, it involves the generation and suggested execution of dynamic code.

resume