The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and installs the firecrawl-cli package from the NPM registry during its setup phase using npm install -g and npx commands.
[REMOTE_CODE_EXECUTION]: The installation instructions include running an initialization script via npx firecrawl-cli init, which fetches and executes code directly from a remote repository. Additionally, the browser command facilitates dynamic code execution through an eval sub-command within a remote browser environment.
[COMMAND_EXECUTION]: The skill is granted permission to run arbitrary firecrawl and npx firecrawl commands via the Bash shell. This includes interacting with a remote Chromium instance where arbitrary JavaScript can be executed.
[PROMPT_INJECTION]: The skill is designed to process untrusted third-party data from the web, creating an attack surface for indirect prompt injection.
Ingestion points: Web content fetched via firecrawl scrape, firecrawl search, and firecrawl crawl as defined in SKILL.md.
Boundary markers: The skill suggests isolating output in a .firecrawl/ directory and utilizing incremental reading techniques (e.g., grep, head) as documented in rules/security.md.
Capability inventory: Broad Bash command execution, file system writes to the .firecrawl/ directory, and JavaScript evaluation (eval) in the browser tool.
Sanitization: The skill lacks automated sanitization of fetched content, relying on manual inspection guidelines.
[CREDENTIALS_UNSAFE]: Authentication procedures documented in rules/install.md involve passing sensitive API keys as command-line arguments (--api-key "<key>"), which can expose them in shell history, logs, or process monitors.

firecrawl