docs(readme): add per-installer and per-OS download badge breakouts

[cmeans-claude-dev]

Summary

• Add per-installer download badges (pip, pipenv, pipx, uv, poetry, pdm) and per-OS download badges (linux, macos, windows) under the existing hero Downloads badge.
• Powered by the per-installer endpoints (pypi-winnow-downloads v0.2.0) and per-OS endpoints (v0.3.0) on pypi-badges.intfar.com.
• Hero Downloads badge unchanged.
• All 10 endpoints verified live before commit.

Test plan

• uv run pytest
• ruff check
• ruff format --check
• mypy
• README renders correctly on GitHub (all 9 new badges visible)

CHANGELOG

• Entry added under ## [Unreleased] → ### Added.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cmeans]

QA Review — PR #99

Verdict: Ready for QA Signoff. Zero findings.

Verification

Gate	Result
uv run pytest	488 passed, 6 deselected (integration marker, expected), 5 xfailed
uv run ruff check src/ tests/	All checks passed
uv run ruff format --check src/ tests/	9 files already formatted
uv run mypy src/mcp_clipboard/	Success: no issues found in 4 source files
Live endpoints (10/10)	All return HTTP 200 with valid shields.io endpoint JSON
Internal arithmetic	per-installer sum = per-OS sum = hero badge count (106) — consistent
CHANGELOG conformance	Under [Unreleased] → ### Added, accurate description
Repo-wide drift grep	README and the prior PR #98 CHANGELOG line are the only files referencing pypi-badges.intfar.com; no drift
CI rollup	All success (lint, typecheck, test 3.11/3.12/3.13, codecov/patch, on-push, qa-approved)

Endpoint sample

Endpoint	Label	Count
installer-pip-30d-non-ci.json	pip (30d)	80
installer-uv-30d-non-ci.json	uv (30d)	26
installer-pipenv/pipx/poetry/pdm	each	0 (lightgrey)
os-linux-30d-non-ci.json	linux (30d)	78
os-macos-30d-non-ci.json	macos (30d)	24
os-windows-30d-non-ci.json	windows (30d)	4
Hero downloads-30d-non-ci.json	pip*/uv/poetry/pdm (30d)	106

pip + uv = 106 and linux + macos + windows = 106 — both breakouts agree with the hero total.

Awaiting maintainer QA Approved.

[cmeans]

Adding Ready for QA Signoff — all gates pass, zero findings on review. CI green, all 10 badge endpoints live and self-consistent (per-installer + per-OS sums match hero total of 106). Awaiting maintainer QA Approved.

[cmeans]

LGTM

[cmeans]

QA Review — Round 2 (head fa2ff87)

Verdict: QA Failed — 1 substantive finding (PR body drift).

Re-verification (all green)

Gate	Result
uv run pytest	488 passed, 6 deselected, 5 xfailed
uv run ruff check src/ tests/	clean
uv run ruff format --check src/ tests/	9 files already formatted
uv run mypy src/mcp_clipboard/	no issues, 4 source files
CI rollup	all SUCCESS (lint, typecheck, test 3.11/3.12/3.13, codecov/patch, on-push)
Endpoints	unchanged from round 1 — all 10 still HTTP 200

Finding

F1 (substantive) — PR body drift after label removal. The round-2 commit fa2ff87 removes the **By installer:** and **By OS:** section labels from the README, but the PR body still references them in two places:

1. Summary, bullet 1: "Add By installer: badge row (pip, pipenv, pipx, uv, poetry, pdm) and By OS: badge row (linux, macos, windows)..." — these row labels no longer exist in the README.
2. Test plan, last item: "README renders correctly on GitHub (badges visible, labels read 'By installer:' and 'By OS:'')" — the labelled-row assertion can no longer be verified because there are no labels.

PR descriptions live forever as historical record once merged, so a present-tense Summary that contradicts the actual diff is worth fixing in the same cycle.

Suggested edits (no new commit needed — just a body edit):

• Bullet 1: drop the **By installer:** / **By OS:** framing, e.g. "Add per-installer download badges (pip, pipenv, pipx, uv, poetry, pdm) and per-OS download badges (linux, macos, windows) under the existing hero Downloads badge."
• Test plan rendering item: rephrase as "README renders correctly on GitHub (all 9 new badges visible)".

CHANGELOG entry is fine as-is — it describes the breakouts conceptually without referencing the bold section labels.

Render observation (not a finding)

With the section labels removed, the three badge groups are now separated only by blank lines. On GitHub render this still produces visible vertical spacing, and the badge labels themselves (pip (30d), linux (30d), …) carry the meaning, so the grouping is recoverable. Not blocking — flagging only because if you reconsider the round-2 simplification, this is the trade-off.

Next step

Fix the PR body, then re-apply Ready for QA.

[cmeans]

Applying QA Failed — round 2 (head fa2ff87) introduced PR body drift: Summary and Test plan still reference **By installer:** / **By OS:** labels that this round removed from the README. Code/docs gates all green; fix is a PR body edit only, no recommit needed. Details in the review above.

[cmeans]

QA Review — Round 3 (head fa2ff87, body edit only)

Verdict: Ready for QA Signoff. F1 from round 2 closed.

F1 (substantive): PR body drift — fixed

Both Summary bullet 1 and the Test plan rendering item now read cleanly:

• Bullet 1: "Add per-installer download badges ... and per-OS download badges ..." — no longer references the removed bold section labels.
• Test plan: "README renders correctly on GitHub (all 9 new badges visible)" — the assertion now matches what's actually in the README.

Re-verification (head unchanged, re-run anyway)

Gate	Result
uv run pytest	488 passed, 6 deselected, 5 xfailed
uv run ruff check src/ tests/	clean
uv run ruff format --check src/ tests/	9 files already formatted
uv run mypy src/mcp_clipboard/	no issues, 4 source files
CI rollup	all SUCCESS

Awaiting maintainer QA Approved.

[cmeans]

Applying Ready for QA Signoff — F1 closed via body edit, all four gates re-verified green on unchanged head fa2ff87. Awaiting maintainer QA Approved.