The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from third-party websites (Amazon, Best Buy, Walmart, etc.), which creates a surface for indirect prompt injection.
Ingestion points: Product pages and search results are fetched using npx webcrawl scrape and npx webcrawl search (SKILL.md).
Boundary markers: The skill contains explicit instructions to the agent to treat fetched content as data rather than instructions and to ignore phrases like "ignore previous instructions" found in web content (SKILL.md).
Capability inventory: The agent can execute CLI commands (npx webcrawl), perform network requests (curl), and read/write local state files (watchlist.json, state.json).
Sanitization: Instructions mandate that the agent requires user approval before taking actions based on fetched content and suggests summarizing data with source URLs.
[PERSISTENCE]: The skill describes patterns for maintaining long-term monitoring through system-level schedulers.
Evidence: Instructions in references/alerting.md guide the setup of cron (Linux), launchd (macOS), or systemd timers to execute the monitoring loop on a schedule.
[COMMAND_EXECUTION]: The skill relies on executing external CLI tools to perform its core functions.
Evidence: Multiple examples in SKILL.md demonstrate the use of npx webcrawl for scraping and data extraction.
[EXTERNAL_DOWNLOADS]: The skill uses npx which may download the webcrawl package from the npm registry if it is not already present in the environment.
Evidence: Usage of npx webcrawl in SKILL.md. The package is a vendor-owned resource belonging to brainfish-ai.

webcrawl-deal-watch