Two files, one repository

u9401066/pubmed-search-mcp ships 3 formats across 21 indexed files. The question worth asking is whether the second one says anything the first does not.

CompareAGENTS.md ↔ Cline rules AGENTS.md ↔ Copilot instructions Cline rules ↔ Copilot instructions

A · AGENTS.md · 412 words B · .clinerules/60-pubmed-python.md · 160 words

What each file covers, counted
Dimension	Shared	Only in A	Only in B	Overlap
Sections	0	6	5	0%
Commands	1	2	4	14%
Section tags	3	1	3	43%

What each file covers

Sections

0 shared · 6 only in A · 5 only in B

− PubMed Search MCP Agent Baseline
− Scope
− Cross-tool Rules
− Important Paths
− Verification Baseline
− Shared Constraints
+ PubMed Python Rules
+ Environment
+ Architecture Guardrails
+ Validation Gates
+ Test Rules

Commands

1 shared · 2 only in A · 4 only in B

− uv run pytest -q
− uv run python scripts/check_async_tests.py
+ uv run ...
+ uv run ruff check .
+ uv run ruff format --check .
+ uv run pytest
uv run mypy src/ tests/

Section tags

3 shared · 1 only in A · 3 only in B

− agent-behaviour
+ setup
+ lint-format
+ testing-strategy
test
code-style
do-not

Line diff

+29 added−54 removed10 unchanged15.6% identical

u9401066/pubmed-search-mcp · AGENTS.md

@@ −1 @@

1# PubMed Search MCP Agent Baseline

3Shared workspace guidance for agent clients that can read `AGENTS.md`.

4Use this file for cross-tool baseline rules. Keep agent-specific behavior in

5tool-native locations such as `.github/copilot-instructions.md` or

6`.clinerules/` so the same instruction is not duplicated in multiple places.

8## Scope

10- This repository is a Python MCP server for biomedical literature search.

11- The architecture follows DDD boundaries: presentation -> application -> domain,

12 with infrastructure isolated behind those layers.

13- Search, timeline, pipeline, export, session, and reference verification are

14 product capabilities. MCP tools should remain thin wrappers over those

15 services.

16- Local literature notes default to wiki-note semantics. Foam-compatible

17 wikilinks, MedPaper-style layouts, CSL JSON, and user templates are export

18 profiles under the application export layer, not presentation-only behavior.

20## Cross-tool Rules

22- Use `uv` for dependency management and `uv run` for Python commands in this repo.

23- Fix root causes instead of layering ad-hoc patches.

24- Preserve DDD separation. Do not move business logic into MCP tool files,

25 hooks, or shell scripts.

26- Keep changes minimal and local. Avoid unrelated refactors when solving a

27 focused problem.

28- Update relevant docs when behavior, tool surface, or setup instructions change.

29- When MCP tools are added, removed, or renamed, keep registry, docs, and

30 generated artifacts in sync.

31- When note export behavior changes, keep skills, Copilot instructions, Cline

32 rules, generated docs, and packaged references aligned.

33- The `unicode-mojibake` pre-commit hook scans newly staged diff lines for

34 corrupted emoji/UTF-8 artifacts; restore UTF-8 text instead of accepting

35 garbled glyphs.

36- Prefer existing repo assets over creating parallel variants. If a shared rule

37 belongs here, do not duplicate it in `.clinerules/` or Copilot-only files.

39## Important Paths

41- `.github/copilot-instructions.md`: GitHub Copilot-specific runtime guidance

42- `.clinerules/`: Cline-specific rules and workflows

43- `.vscode/mcp.json`: workspace MCP setup for VS Code / Copilot Chat

44- `.claude/skills/pipeline-persistence/references/`: packaged tutorial copies

45 for agent bundles and VSIX integrations that do not read `docs/site-content/`

46- `scripts/setup-vscode-ai-harness.sh`: install recommended VS Code extensions

47- `docs/INTEGRATIONS.md`: client configuration reference

49## Verification Baseline

51- Narrow changes: run the smallest relevant validation first.

52- Repo-wide validation when integration surfaces change:

53 - `uv run pytest -q`

54 - `uv run mypy src/ tests/`

55 - `uv run python scripts/check_async_tests.py`

57## Shared Constraints

59- Do not add duplicate rule sets across tools. Put shared baseline rules here.

60- Put Cline-only workflow automation in `.clinerules/workflows/`.

61- Put Copilot-only hooks/chatmode/instruction behavior under `.github/`.

62- Keep workspace-level AI assets version-controlled so users get a usable setup

63 immediately after clone or packaging.

u9401066/pubmed-search-mcp · .clinerules/60-pubmed-python.md

@@ +1 @@

1---

2paths:

3 - "src/**/*.py"

4 - "tests/**/*.py"

5 - "scripts/**/*.py"

6 - "pyproject.toml"

7 - "uv.lock"

8---

10# PubMed Python Rules

12## Environment

14- Use `uv` for all Python commands.

15- Prefer `uv run ...` so checks execute in the right environment.

16- Do not introduce raw package-manager calls in docs, CI, or code paths.

18## Architecture Guardrails

20- Keep `domain/` free of I/O, environment reads, and framework imports.

21- Keep external API calls in `infrastructure/`.

22- Put orchestration in `application/`.

23- Keep MCP presentation code as thin wrappers around services.

24- Read configuration through shared settings, not scattered `os.environ` reads.

26## Validation Gates

28- Lint: `uv run ruff check .`

29- Format: `uv run ruff format --check .`

30- Types: `uv run mypy src/ tests/`

31- Tests: `uv run pytest`

33## Test Rules

35- Use async tests and `AsyncMock` for async methods.

36- Preserve xdist-friendly tests; avoid singleton state leaks.

37- Mock external APIs unless a test is explicitly marked integration.

38- Add regression tests for parsing, query expansion, ranking, source counts, and MCP tool contracts.

@@ −1 +1 @@

1−# PubMed Search MCP Agent Baseline

1+---

2+paths:

3+ - "src/**/*.py"

4+ - "tests/**/*.py"

5+ - "scripts/**/*.py"

6+ - "pyproject.toml"

7+ - "uv.lock"

8+---

3−Shared workspace guidance for agent clients that can read `AGENTS.md`.

4−Use this file for cross-tool baseline rules. Keep agent-specific behavior in

5−tool-native locations such as `.github/copilot-instructions.md` or

6−`.clinerules/` so the same instruction is not duplicated in multiple places.

10+# PubMed Python Rules

711

8−## Scope

12+## Environment

913

10−- This repository is a Python MCP server for biomedical literature search.

11−- The architecture follows DDD boundaries: presentation -> application -> domain,

12− with infrastructure isolated behind those layers.

13−- Search, timeline, pipeline, export, session, and reference verification are

14− product capabilities. MCP tools should remain thin wrappers over those

15− services.

16−- Local literature notes default to wiki-note semantics. Foam-compatible

17− wikilinks, MedPaper-style layouts, CSL JSON, and user templates are export

18− profiles under the application export layer, not presentation-only behavior.

14+- Use `uv` for all Python commands.

15+- Prefer `uv run ...` so checks execute in the right environment.

16+- Do not introduce raw package-manager calls in docs, CI, or code paths.

1917

20−## Cross-tool Rules

18+## Architecture Guardrails

2119

22−- Use `uv` for dependency management and `uv run` for Python commands in this repo.

23−- Fix root causes instead of layering ad-hoc patches.

24−- Preserve DDD separation. Do not move business logic into MCP tool files,

25− hooks, or shell scripts.

26−- Keep changes minimal and local. Avoid unrelated refactors when solving a

27− focused problem.

28−- Update relevant docs when behavior, tool surface, or setup instructions change.

29−- When MCP tools are added, removed, or renamed, keep registry, docs, and

30− generated artifacts in sync.

31−- When note export behavior changes, keep skills, Copilot instructions, Cline

32− rules, generated docs, and packaged references aligned.

33−- The `unicode-mojibake` pre-commit hook scans newly staged diff lines for

34− corrupted emoji/UTF-8 artifacts; restore UTF-8 text instead of accepting

35− garbled glyphs.

36−- Prefer existing repo assets over creating parallel variants. If a shared rule

37− belongs here, do not duplicate it in `.clinerules/` or Copilot-only files.

20+- Keep `domain/` free of I/O, environment reads, and framework imports.

21+- Keep external API calls in `infrastructure/`.

22+- Put orchestration in `application/`.

23+- Keep MCP presentation code as thin wrappers around services.

24+- Read configuration through shared settings, not scattered `os.environ` reads.

3825

39−## Important Paths

26+## Validation Gates

4027

41−- `.github/copilot-instructions.md`: GitHub Copilot-specific runtime guidance

42−- `.clinerules/`: Cline-specific rules and workflows

43−- `.vscode/mcp.json`: workspace MCP setup for VS Code / Copilot Chat

44−- `.claude/skills/pipeline-persistence/references/`: packaged tutorial copies

45− for agent bundles and VSIX integrations that do not read `docs/site-content/`

46−- `scripts/setup-vscode-ai-harness.sh`: install recommended VS Code extensions

47−- `docs/INTEGRATIONS.md`: client configuration reference

28+- Lint: `uv run ruff check .`

29+- Format: `uv run ruff format --check .`

30+- Types: `uv run mypy src/ tests/`

31+- Tests: `uv run pytest`

4832

49−## Verification Baseline

33+## Test Rules

5034

51−- Narrow changes: run the smallest relevant validation first.

52−- Repo-wide validation when integration surfaces change:

53− - `uv run pytest -q`

54− - `uv run mypy src/ tests/`

55− - `uv run python scripts/check_async_tests.py`

56−

57−## Shared Constraints

58−

59−- Do not add duplicate rule sets across tools. Put shared baseline rules here.

60−- Put Cline-only workflow automation in `.clinerules/workflows/`.

61−- Put Copilot-only hooks/chatmode/instruction behavior under `.github/`.

62−- Keep workspace-level AI assets version-controlled so users get a usable setup

63− immediately after clone or packaging.

35+- Use async tests and `AsyncMock` for async methods.

36+- Preserve xdist-friendly tests; avoid singleton state leaks.

37+- Mock external APIs unless a test is explicitly marked integration.

38+- Add regression tests for parsing, query expansion, ranking, source counts, and MCP tool contracts.

6439