The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the facebook-marketplace-pp-cli tool and related modules from official vendor sources.
Fetches the CLI package via the @mvanhorn/printing-press-library NPM package.
Downloads and installs source code from the author's GitHub repository at github.com/mvanhorn/printing-press-library using the Go toolchain.
[COMMAND_EXECUTION]: The skill is designed around the execution of the facebook-marketplace-pp-cli binary to perform various marketplace operations.
Uses npx to run installation commands.
Invokes shell commands for searching, watching, drafting, and replying to listings.
[DATA_EXFILTRATION]: The tool includes a --deliver webhook:<url> flag which allows the agent to route command outputs, including sensitive marketplace data or inbox content, to arbitrary external URLs.
Additionally supports an --deliver file:<path> sink for writing output to the local file system.
Environment variables like FACEBOOK_MARKET_FEEDBACK_ENDPOINT can be configured to automatically send local feedback logs to a remote server.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external Facebook Marketplace users, which could contain malicious instructions.
Ingestion points: Untrusted data enters the context via inbox thread (buyer messages), marketplace_search (listing titles/descriptions), and listing get (listing metadata) in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded in the retrieved marketplace content.
Capability inventory: The skill uses facebook-marketplace-pp-cli which has capabilities for file writing (--deliver file), network POSTing (--deliver webhook), and automated messaging (reply).
Sanitization: There is no mention of sanitization, validation, or escaping of the external content before it is processed by the agent or the CLI.

pp-facebook-marketplace