The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (cat and python3) to generate Word documents. The workflow involves writing content to /tmp/lit_review_temp.md and executing a local script scripts/generate_docx.py. While functional, the inclusion of user-provided content like the paper title in a shell command string (--title "[论文标题]") presents a potential command injection surface if the agent does not properly escape shell metacharacters.
[INDIRECT_PROMPT_INJECTION]: The skill systematically ingests data from external sources via WebSearch (Google Scholar, ScienceDirect, etc.) to compile literature reviews. Malicious instructions or deceptive content embedded in indexed academic papers or web snippets could potentially influence the agent's output. The skill includes a mandatory verification step (literature-verifier) and quality checklist which act as mitigations.
Ingestion points: Step 2 in SKILL.md uses WebSearch to fetch external literature data.
Boundary markers: None explicitly defined for the search results beyond the verification-guide.md instructions.
Capability inventory: The skill can write files to /tmp and ~/Desktop, and execute Python scripts.
Sanitization: Step 3 (literature-verifier) is designed to verify the existence and metadata of papers, acting as a filter for AI-generated hallucinations or potentially unreliable sources.
[EXTERNAL_DOWNLOADS]: The Python script scripts/generate_docx.py relies on the python-docx library. While the skill does not explicitly show the installation command, this is an external dependency required for the tool to function.

literature-review-writer