[MarkusNeusinger]

Summary

• Owner decision 2026-10-15: the CLAUDE.md rule "fix small build/test blockers directly, even when out of scope" applies to every agent. .github/copilot-instructions.md gets it under Important Rules, with the under-~20-lines bar and ask-first above it. Its Common Issues line used to tell Copilot to test only its own changes and leave pre-existing failures alone; it now points at the rule.
• Owner decision 2026-10-15: Python tooling is uv only. The two pip fallbacks in the Copilot guide are gone, and CLAUDE.md "Package Management" says so explicitly so the rule is mirrored.
• One more clear contradiction, found while comparing the two guides: Task Suitability listed "adding new plot types following existing patterns" as a good Copilot task, which the mandatory spec/impl pipeline forbids. It now says to propose them as a spec-request issue.
• tests/unit/test_agent_instructions.py pins both rules as mirrored. The keywords carry the obligation and its limit, not only the subject.

Other differences between the two guides are listed in the PR conversation, not fixed here; they are omissions or owner calls rather than contradictions.

Plan

N/A

Test plan

• uv run pytest tests/unit/test_agent_instructions.py (34 passed)
• uv run ruff check and uv run ruff format --check on the test file
• uv run python -m tools.changelog check --base origin/main

[MarkusNeusinger]

Remaining differences between CLAUDE.md and .github/copilot-instructions.md, compared on 2026-10-15. None of them is a contradiction, so this PR leaves them alone:

• Rules only in CLAUDE.md: "Proper lint fixes only" (no noqa or eslint-disable without approval), "Modify repo files only with Edit/Write", "Snapshot before destructive prod operations", and "Update documentation when making changes". Copilot isn't told otherwise, but it isn't told these either. Mirroring "Proper lint fixes only" would help most, because the Copilot guide lists "Fixing linting/formatting issues" as a good Copilot task.
• Copilot Task Suitability: "Bug fixes in existing plot implementations". CLAUDE.md doesn't forbid hand edits to plots/*/implementations/, but the regen and update flows normally own them.
• Copilot plot style example: it hardcodes color='#306998' and figsize=(16, 9). That example is about the KISS structure, not the palette or canvas, which prompts/ govern. Reconciling it is a catalogue-style question, not an instruction conflict.

[Copilot]

🟢 Approval recommended

The focused guidance updates and matching test assertions have no unresolved blocking issues.

0 open findings

What changed in this PR

Aligns Copilot guidance with CLAUDE.md and adds tests to prevent shared rules from drifting.

Changes:

• Aligns small-blocker fixes and uv-only Python tooling guidance.
• Routes new plot proposals through spec-request issues.
• Adds mirrored-rule assertions and a changelog fragment.

File	Description
tests/​unit/​test_agent_instructions.py	Pins both shared rules.
CLAUDE.md	Makes uv-only tooling explicit.
changelog.d/​align-copilot-instructions.md	Records the guidance alignment.
.github/​copilot-instructions.md	Resolves conflicting repair, tooling, and plot-proposal guidance.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!