The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install external binaries and packages from vendor-owned sources.
NPM package installation: npx -y @mvanhorn/printing-press-library install human-goat --cli-only.
Go module installation: go install github.com/mvanhorn/printing-press-library/library/productivity/human-goat/cmd/human-goat-pp-cli@latest.
[COMMAND_EXECUTION]: The skill's primary function is the execution of the human-goat-pp-cli binary with various subcommands (hire, cancel, best, spend, etc.) to perform real-world tasks.
[DATA_EXFILTRATION]: The skill documents an output delivery system using the --deliver webhook:<url> flag, which allows the agent to POST command results (potentially containing account details or personal task information) to arbitrary external URLs.
[DYNAMIC_EXECUTION]: The skill implements an 'Automatic learning' loop that synthesizes and executes 'playbooks' (pre-defined sequences of CLI commands) stored in local JSON files. These playbooks are generated from previous session journals and replayed during future invocations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user queries and external human task networks (TaskRabbit and Magic) which is then used to synthesize playbooks or influence autonomous hiring decisions.
Ingestion points: User input for the hire, dispatch, recall, and teach commands; task status and descriptions fetched from external human networks.
Boundary markers: None identified in the skill instructions to separate processed data from command logic.
Capability inventory: Execution of shell commands via the CLI, writing to local learning stores via teach, and network exfiltration via the webhook sink.
Sanitization: The skill includes instructions for the agent to strip PII (emails, phone numbers) before teaching new commands, though this relies on agent compliance rather than hard technical constraints.

pp-human-goat