Pharmaclaw Literature Agent
PassAudited by VirusTotal on May 11, 2026.
Overview
Type: OpenClaw Skill Name: pharmaclaw-literature-agent Version: 2.0.0 The OpenClaw AgentSkills bundle 'pharmaclaw-literature-agent' is benign. It functions as a literature mining agent, integrating searches across public APIs for PubMed (NCBI E-utilities), Semantic Scholar, ClinicalTrials.gov, and bioRxiv/medRxiv. All network requests are handled by the standard `requests` library, which properly encodes query parameters, mitigating command injection risks. The `SKILL.md` documentation is straightforward, accurately describing the agent's purpose and usage without any prompt injection attempts. No evidence of data exfiltration, malicious execution, persistence mechanisms, or obfuscation was found across any of the Python scripts (`biorxiv_search.py`, `chain_entry.py`, `chain_entry_v2.py`, `clinicaltrials_search.py`, `pubmed_search.py`, `semantic_scholar.py`).
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Search terms such as unpublished compound names, target strategies, or disease programs could be visible to external API providers.
The skill clearly discloses that it sends literature queries to external public providers; this is purpose-aligned but means proprietary or sensitive research terms may be shared with those services.
Searches PubMed (NCBI E-utilities) and Semantic Scholar ... ClinicalTrials Phase II/III + bioRxiv preprints. No API keys required.
Avoid entering confidential or proprietary research terms unless you are comfortable sending them to the named public literature services.
Returned literature content can influence later agent reasoning or recommendations, so incorrect or adversarial external text could affect conclusions.
The skill brings external article abstracts, summaries, and metadata into the agent context and may use them to suggest downstream workflows.
Returns structured results with titles, authors, abstracts, DOIs, MeSH terms, citation counts, TLDR summaries... Chains from any PharmaClaw agent ... and recommends next agents based on findings.
Treat summaries and abstracts as untrusted research evidence; verify important claims against the cited sources before acting on them.
The skill may fail until the expected Python dependency is present, and users have less formal dependency/provenance information than a pinned install spec would provide.
The included scripts depend on the requests package, while the supplied install section declares no install spec; this is a small dependency documentation gap.
try:
import requests
except ImportError:
print(json.dumps({"status": "error", "error": "requests not installed"}))Install dependencies from trusted sources and prefer a reviewed or pinned dependency setup if using this skill operationally.