The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites (retailers, coupon aggregators, and social media) to extract pricing and seller information.
Ingestion points: External content enters the agent context via the scripts/firecrawl.sh wrapper or the agent's native fetch capabilities.
Boundary markers: The instructions do not define explicit boundary markers or "ignore" instructions for the scraped content before the agent processes it.
Capability inventory: The skill utilizes shell execution (via scripts/firecrawl.sh) and Docker management for the Firecrawl backend.
Sanitization: No specific sanitization or filtering of the external markdown/text is described in the 4-step pipeline.
[EXTERNAL_DOWNLOADS]: The references/self-host-firecrawl/docker-compose.yml file references several container images from established organizations.
Sources: It downloads images from the GitHub Container Registry (ghcr.io/firecrawl/*) and official Docker Hub repositories for Redis, RabbitMQ, and SearXNG.
Integrity: All service images are pinned to verified SHA256 digests to ensure reproducibility and protect against supply chain tampering.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary workflows.
Evidence: The core logic involves running scripts/firecrawl.sh and Docker commands for backend setup.
Safety: The firecrawl.sh script includes a json_str function that safely encodes user-provided search queries and URLs using jq, python3, or sed before they are interpolated into the API request body, preventing command injection.

shop-scout