The Agent Skills Directory

[SAFE]: The skill operates entirely within the scope of academic text generation and synthesis. It does not invoke any shell commands, request network access, or attempt to access sensitive system files. All file operations are restricted to reading local project artifacts (e.g., research-lineage.md, figure-table-list.md) for context.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of user-supplied manuscripts and metadata artifacts. This establishes a theoretical attack surface for indirect prompt injection, though the lack of high-privilege tools or exfiltration pathways renders this risk minimal. 1. Ingestion points: SKILL.md and static/core/workflow.md define processes for reading full thesis manuscripts and thesis-context artifacts from the workspace. 2. Boundary markers: The instructions utilize [TBD: ...] placeholders for missing content but lack explicit delimiters (e.g., XML tags or unique markers) to isolate untrusted user data from the agent's core instructions. 3. Capability inventory: The skill does not possess capabilities for subprocess execution, file writing, or external network communication. 4. Sanitization: No explicit content filtering or sanitization of external text is defined in the workflow.

cn-thesis-drafting