The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through its processing of external content. Instructions embedded in a job description or a resume file could influence the agent's behavior during the customization process.
Ingestion points: The skill ingests untrusted data via the job.parse <url> command and the resume.parse <file> workflow defined in SKILL.md and references/commands.md.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands when processing external text, increasing the likelihood that the model might follow malicious instructions hidden in a job posting.
Capability inventory: The agent can export files (resume.export), list variants, and delete user data (gdpr.delete), though no high-privilege system command execution was detected.
Sanitization: While references/hooks.md defines validation for file types and sizes, there is no mention of sanitizing the natural language content for injection patterns or escaping external strings before interpolation.
[EXTERNAL_DOWNLOADS]: The job.parse command (found in references/commands.md) allows the agent to fetch content from user-provided URLs. This is a functional requirement for analyzing job postings but involves networking with external, unverified domains.

resume-writing