The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation instructs users to install the @actalk/inkos package from the npm registry. The @actalk scope is not recognized as a known trusted organization or a verified vendor resource associated with the author 'reason-machines'.
[INDIRECT_PROMPT_INJECTION]: The multi-agent novel writing system ingests untrusted text data from user-controlled files, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Files such as story/outline.md, story/book_rules.md, and external reference novel texts are read by the ArchitectAgent and WriterAgent to influence the writing and auditing process.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings for the data being processed by the LLM-based agents.
Capability inventory: The skill has the capability to perform network operations (LLM APIs, webhooks) and modify the local file system (writing, revising, and deleting chapter files).
Sanitization: There is no evidence of input validation or sanitization to prevent adversarial instructions within the processed text from overriding agent behavior.
[COMMAND_EXECUTION]: The core functionality of the skill involves executing various inkos CLI commands (e.g., inkos book create, inkos write next, inkos daemon), which run logic contained in the externally installed package.
[PRIVILEGE_ESCALATION]: The recommended installation method (npm install -g) typically requires elevated administrative or root privileges, increasing the potential impact if the external package were to be compromised.

inkos-multi-agent-novel-writing