The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown instructions and a methodological guide (references/reference-originality-method.md). It does not contain any executable scripts, shell commands, or tool configurations that would permit system access or automated operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external 'subjects' and 'references' provided by the user. While this represents a vulnerability surface, the risk is considered safe based on the following evidence: 1. Ingestion points: The agent processes external products, revisions, and artifacts as described in the comparison registry. 2. Boundary markers: The methodology uses structured logging templates and specific record fields (Pair ID, Subject locator) to delimit evidence. 3. Capability inventory: The skill has no allowed tools, subprocess calls, or network operations, preventing the execution of any instructions found in the untrusted data. 4. Sanitization: No specific filtering is described, but the lack of executable capabilities eliminates the potential for exploitation.

reference-originality