Cline rules

1# PubMed Search MCP Project Rules
2 
3These rules are meant for Cline usage in the PubMed Search MCP repository.
4 
5## Goals
6 
7- Provide a robust MCP server for biomedical literature search, analysis, full-text access, and citation export.
8- Make `unified_search` the default entrypoint while preserving specialized tools for PICO, MeSH, full text, citations, and pipelines.
9- Keep outputs useful for downstream agents and for Zotero Keeper imports.
10 
11## Repo Layout
12 
13- `src/pubmed_search/domain/`: domain entities and value objects.
14- `src/pubmed_search/application/`: search, export, pipeline, and analysis use cases.
15- `src/pubmed_search/infrastructure/`: external clients and source integrations.
16- `src/pubmed_search/presentation/mcp_server/`: MCP server, tools, prompts, and resources.
17- `tests/`: pytest suite.
18- `.claude/skills/`: user-facing research skills.
19- `.cline/skills/` and `.clinerules/`: Cline harness assets.
20 
21## Canonical Commands
22 
23- `uv run ruff check .`
24- `uv run ruff format --check .`
25- `uv run mypy src/ tests/`
26- `uv run pytest`
27- `uv run python scripts/count_mcp_tools.py --update-docs`
28- `python3 scripts/check_cline_skills.py`
29 
30## Safety / Hygiene
31 
32- Avoid editing generated outputs unless the generator was run intentionally.
33- Never commit NCBI API keys, Semantic Scholar keys, CORE keys, OpenAlex keys, cookies, or institutional resolver secrets.
34- Avoid destructive git operations unless explicitly requested.
35- Keep network-dependent behavior mockable and rate-limit aware.
36 
37## Prefer Existing Patterns
38 
39- Reuse source adapters and registry abstractions instead of adding one-off HTTP calls.
40- Keep MCP tools small and delegate behavior to application services.
41- Keep backward-compatible aliases where agents already consume fields.
42- Update tool docs and user-facing skills when tool contracts change.
43