The Agent Skills Directory

[SAFE]: The skill is composed of natural language instructions and markdown formatting without any executable scripts, binaries, or automated command triggers.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill requests 'file:read' permission to access manuscript files. This permission is aligned with the skill's purpose and no indicators of credential harvesting or data exfiltration to external domains were identified.
[INDIRECT_PROMPT_INJECTION]: The skill processes manuscript text from external files which constitutes an ingestion point for untrusted data. 1. Ingestion points: Manuscript content read from local files. 2. Boundary markers: No explicit markers are defined, although the skill instructs the agent to use specific quotes. 3. Capability inventory: Restricted to 'file:read'; no network, subprocess, or execution capabilities are requested. 4. Sanitization: None identified. The risk of malicious instruction execution is minimal due to the skill's narrow operational scope and lack of high-privilege tools.

continuity-check