The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted job descriptions from Upwork (via the Apify scraper) and interpolates this content into LLM prompts in scripts/upwork_proposal_generator.py. This represents a potential indirect prompt injection surface where a malicious job listing could attempt to influence the proposal generation process. However, this is inherent to the skill's primary purpose of processing external job data.
Ingestion points: Reads from .tmp/upwork_jobs.json containing scraped Upwork data.
Boundary markers: Absent; data is directly interpolated into f-strings.
Capability inventory: Accesses Anthropic API, Google Drive API (document creation), and Google Sheets API.
Sanitization: None performed on job titles or descriptions before LLM processing.
[DATA_EXPOSURE]: The scripts scripts/update_sheet.py and scripts/upwork_proposal_generator.py access local credential files (token.json, service_account.json) to authenticate with Google APIs. This is standard secret management for Google Cloud integration and is not an exfiltration risk in this context.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to well-known service endpoints including api.apify.com, api.anthropic.com, and googleapis.com. These operations are required for the skill's documented functionality and target trusted providers.

upwork-apply