The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides an environment setup module (skills/environment-setup/SKILL.md) that downloads the Miniconda installer from repo.anaconda.com. This is a well-known service and the download is essential for the skill's primary functionality of managing Python environments for data analysis and visualization.
[COMMAND_EXECUTION]: The skill executes several local utility scripts to initialize project structures (scripts/init_plan.sh), perform academic style checks (scripts/style_check.sh), and parse PDF documents (scripts/pdf_parser.py). These operations are confined to the user's project directory and are transparently documented.
[DATA_EXFILTRATION]: The literature review module fetches metadata and abstracts from established academic services, including PubMed (ncbi.nlm.nih.gov), CrossRef (api.crossref.org), Semantic Scholar (api.semanticscholar.org), and arXiv (arxiv.org). These connections are used solely for retrieving academic information requested by the user.
[PROMPT_INJECTION]: The skill processes data from external search results and user-provided PDFs, which introduces a surface for indirect prompt injection. This is mitigated by the verification skill and peer-review guidelines that instruct the agent to verify all claims and citations against authoritative sources.
[DYNAMIC_EXECUTION]: The skill uses a hook mechanism (hooks/session-start) to inject context into the agent session upon startup. This script simply reads the main instructions to ensure the agent follows the defined research workflow and does not execute untrusted external code.

research-writing-assistant