The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes metadata status: reviewed. As per standard security practices, self-declared safety or review statuses in skill metadata are treated as claims and not as objective proof of safety.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its core function of processing user-provided manuscript chapters.\n
Ingestion points: Manuscript text is ingested via user input or fetched from files using the Model Context Protocol (MCP) as described in the intake and MCP sections of SKILL.md.\n
Boundary markers: The instructions do not specify explicit delimiters or 'ignore' instructions to wrap the manuscript content, which is a standard design for such tools.\n
Capability inventory: The skill has the ability to write back to files via MCP by 'pinning marginalia,' though it is explicitly restricted by its instructions from rewriting original prose.\n
Sanitization: No specific input sanitization or filtering is described for the manuscript content.\n- [SAFE]: The skill references Calliope (MCP), a tool associated with the author calliope-editor. This integration is a documented feature for providing feedback within the author's supported environment.

beta-reader-panel