The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a custom binary named amc-theatres-pp-cli and an MCP server amc-theatres-pp-mcp via the Bash tool to perform theatre and showtime queries.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install external software from sources associated with the author:\n
The @mvanhorn/printing-press-library package via npx.\n
The amc-theatres-pp-cli Go module from github.com/mvanhorn/printing-press-library.\n- [INDIRECT_PROMPT_INJECTION]: The skill implements a self-capturing learning loop (recall, teach, playbook amend) that processes user-controlled input stored in the $USER_QUESTION environment variable. This allows the tool to 'learn' from previous interactions and potentially store malicious instructions if they are included in user queries.\n
Ingestion points: User input provided via the $USER_QUESTION environment variable to the recall, teach, and playbook amend commands in SKILL.md.\n
Boundary markers: The skill documentation notes that standard shell quoting is insufficient to prevent command substitution from user-controlled text.\n
Capability inventory: The skill possesses the Bash tool and can execute the amc-theatres-pp-cli binary, which has capabilities for file system access and network delivery via webhooks.\n
Sanitization: The instructions advise the user to strip personally identifiable information (PII) before teaching, but there is no automated sanitization for prompt injection patterns.\n- [DYNAMIC_EXECUTION]: The skill uses a 'playbook' system where sequences of commands are stored in JSON files and replayed by the agent. These playbooks are synthesized and amended based on session journals, which represents a form of dynamic execution guided by processed user data.

pp-amc-theatres