The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted job description data from external websites, creating a surface for indirect prompt injection. However, it implements mandatory evidence-based mitigations:
Ingestion points: Job descriptions are pulled from Apify datasets and processed by sub-agents as described in reference/fit-scoring.md.
Boundary markers: The skill uses <job_description> and <candidate_profile> XML-style delimiters to isolate untrusted content from system instructions.
Capability inventory: Capabilities are limited to local file operations and API calls to the Apify platform.
Sanitization: Explicit system instructions are provided to the sub-agent to ignore and report any natural language instructions found within the scraped job descriptions, treating them as low-quality metadata rather than valid commands.
[EXTERNAL_DOWNLOADS]: The skill orchestrates the execution of third-party Apify Actors (agentx/all-jobs-scraper, misceres/indeed-scraper, memo23/glassdoor-scraper-ppr). It provides full transparency regarding these dependencies, including their source, functionality, and cost implications. It also implements a confirmation threshold ($20) to prevent unexpected charges.
[COMMAND_EXECUTION]: Utilizes the Apify CLI and curl for interaction with the Apify platform. It follows security best practices for secret management, specifically instructing users to pass API tokens via HTTP headers instead of URL parameters to avoid exposure in system logs.

apify-jobs-data