The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains explicit instructions to override standard safety protocols (the 'Collaboration Protocol'). It directs the agent to allow sub-agents to write files without individual user approval, defining a 'bounded exception' to reduce operational friction. This instruction actively seeks to bypass the standard confirmation requirements for file modifications.- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !bash syntax within its documentation to execute a local shell script (yaml-helper.sh) at load time. This is used to dynamically resolve configuration keys like review_mode and automation before user interaction occurs.- [INDIRECT_PROMPT_INJECTION]: The orchestrator is designed to read and process external narrative assets, including story bibles and character profiles, which serve as ingestion surfaces for potentially malicious instructions. 1. Ingestion points: Reads files from design/narrative/ and character/faction definitions. 2. Boundary markers: The skill does not implement clear delimiters or instructions to ignore potential commands embedded within the story data. 3. Capability inventory: The agent has access to Write, Edit, Bash, and Agent tools, which could be leveraged if an indirect injection were successful. 4. Sanitization: No explicit sanitization or validation of the input narrative text is performed before it is passed to sub-agents.- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute a configuration helper script. The access is restricted to a specific path relative to the skill directory via the allowed-tools configuration.

team-narrative