| name | pubmed-search-mcp-harness |
| description | Codex harness for PubMed Search MCP. Triggers: pubmed search, literature search, unified_search, pipeline, fulltext, release checklist, Codex. |
PubMed Search MCP: Codex Harness Skill
Use this skill when working with Codex on PubMed Search MCP research workflows,
pipeline state, fulltext access, exports, and release checks.
What To Read First
AGENTS.md for Codex workspace instructions.
.github/agents/research.agent.md for research-agent behavior.
.github/hooks/ and scripts/hooks/copilot/ for pipeline enforcement assets.
.claude/skills/pubmed-* and .claude/skills/pipeline-persistence for user-facing workflows.
.clinerules/ for project/release rules that also apply to Codex.
Canonical Commands
- Lint:
uv run ruff check .
- Format check:
uv run ruff format --check .
- Types:
uv run mypy src/ tests/
- Tests:
uv run pytest
- Tool docs sync:
uv run python scripts/count_mcp_tools.py --update-docs
- Skill audit:
python scripts/check_cline_skills.py
Research Guardrails
- Prefer
unified_search as the primary multi-source entrypoint.
- For external Python package/API questions, prefer the stable
pubmed_search.api facade over MCP presentation modules.
- Preserve session/pipeline state instead of relying on agent memory.
- Respect NCBI email/API-key policy and institutional access configuration.
- Keep exported citations reproducible with PMIDs, DOI, source, date filters, and query strategy.