The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted job posting content from web searches and local files, creating a surface for indirect instructions to influence the agent. * Ingestion points: Job postings retrieved via WebFetch and stored job data in job_search_tracker.csv and seen_jobs.json. * Boundary markers: Present in the form of instructions (Rule 8) telling the agent to treat stored gaps as data and not to follow embedded instructions, though no structural delimiters are used. * Capability inventory: Read, Write, WebFetch, WebSearch. * Sanitization: No sanitization of external text is specified before it is used in the synthesis of the learning plan.
[DATA_EXFILTRATION]: The skill accesses local files containing sensitive professional and personal data. * Evidence: Accesses .claude/skills/job-application-assistant/01-candidate-profile.md and job_search_tracker.csv to compare the user's current skills against job requirements.

upskill