The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources, which could contain malicious instructions designed to influence the agent's behavior during the cover-letter generation process.
Ingestion points: The skill ingests text from CV/resume files (PDF, Markdown, LaTeX, TXT) and job descriptions (pasted text or file paths) in Step 0 and Step 1. It also performs a recursive repository scan for supplementary context.
Boundary markers: There are no specific delimiters or "ignore embedded instructions" markers defined to separate the untrusted data from the system prompts.
Capability inventory: The skill possesses the ability to execute shell commands (pdftotext, sed, pandoc) and write new Markdown files to the local filesystem.
Sanitization: The instructions do not specify any validation, escaping, or filtering of the extracted text before it is synthesized into the profile used for writing the cover letter.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell utilities to extract and transform text from files.
Evidence: In Step 1, the skill uses pdftotext for PDF extraction, sed for regex-based cleaning of LaTeX files, and pandoc for document conversion. While these are legitimate uses for the skill's primary purpose, they involve processing potentially complex file formats provided by users.

cover-letter-builder