The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes search and crawling tools (such as Argo or host-provided search/browser capabilities) to perform fact-checking and evidence verification. This is limited to the skill's primary function of information retrieval.
[EXTERNAL_DOWNLOADS]: Performs network operations to access public documents, standards, and literature for the purpose of verifying claims in the user-provided text. This behavior is necessary for its stated fact-checking purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (user documents). It addresses this risk by instructing the agent to maintain a skeptical posture ("assume it might be AI-generated"), prioritize cross-referencing, and report when information cannot be verified, which helps mitigate potential instruction injection within the audited content.
Ingestion points: Processes user-provided documents, reports, and citations in the main prompt context.
Boundary markers: Uses a specific structured output format ("审稿评价报告") to separate analysis from the input data.
Capability inventory: Limited to tool-based search and crawling; no file-writing or arbitrary command execution described.
Sanitization: Focuses on verification and reporting rather than direct execution of content found in documents.

fact-review