fix(install): stamp .graphify_version only for the platform being installed (#2694)

[ousamabenyounes]

Fix #2694 (remaining version-stamp half; the CLAUDE_CONFIG_DIR half landed in v0.9.44).

Problem

graphify install --platform X calls _refresh_all_version_stamps(), which writes the current __version__ into .graphify_version for every already-installed platform's skill dir — including platforms whose SKILL.md this run never rewrote. Their content is still stale, but the fresh stamp makes _check_skill_version skip the skill is from graphify X, package is Y. Run 'graphify install' to update. warning. The stamp asserts a refresh that never happened, so the user is never prompted to update the genuinely-stale skill.

Fix

Each platform's own install path (_copy_skill_file) already writes that platform's .graphify_version when it (re)writes the skill content. So the cross-platform refresh is pure over-stamping: remove _refresh_all_version_stamps(), its call in install(), and its re-export. A platform's stamp now advances only when its own content is (re)written, keeping the staleness warning truthful. This is the direction the issue requests: "Only stamp .graphify_version for platforms whose skill content was actually (re)written in this run; leave other platforms' stamps untouched."

Test verification (RED → GREEN)

New focused regression test tests/test_install_version_stamp.py: seed codex as a previously-installed platform stamped at an old version, then install("claude"), and assert codex's stamp is untouched while claude's is current.

RED (on unmodified v8, before the fix):

E   assert '0.9.44' == '0.0.1-old'
FAILED tests/test_install_version_stamp.py::test_install_does_not_bump_other_platforms_stamp

GREEN (with the fix):

tests/test_install_version_stamp.py::test_install_does_not_bump_other_platforms_stamp PASSED
1 passed

Full local suite

• uv run --frozen pytest tests/ -q → 4514 passed, +1 new test; the only 3 failures (tests/test_ollama.py::test_detect_backend_*) are pre-existing on unmodified v8 too — a local GEMINI_API_KEY env leak, unrelated to this change (they pass with the var unset).
• skillgen --check / --audit-coverage / --always-on-roundtrip: OK (no generated files touched).
• Install/uninstall round-trip suites: green.

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

This PR removes the _refresh_all_version_stamps helper from graphify/install.py (and its re-export in __main__.py), which previously bumped the .graphify_version stamp of all other already-installed platforms after a global install. As a result, install no longer touches version stamps for platforms it didn't re-write during a given run. It also adds a new regression test (tests/test_install_version_stamp.py) covering that installing one platform leaves another platform's stamp untouched, plus a CHANGELOG entry describing the change.

No blocking issues surfaced. 1 lower-confidence candidate did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 572 functions depend on the 342 functions this change touches.

Health — this change adds coupling hotspots:

• new: dispatch_command() — 2 callers, 117 callees
• new: codebuddy_install() — 20 callers, 5 callees
• new: claude_install() — 19 callers, 4 callees
• new: gemini_install() — 10 callers, 7 callees
• new: claude_uninstall() — 17 callers, 4 callees
• new: _project_uninstall() — 5 callers, 13 callees
• new: dispatch_install_cli() — 2 callers, 31 callees
• new: codebuddy_uninstall() — 14 callers, 4 callees
• …and 13 more

Verification — 572 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 572 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify install.

The verifier did not have enough to check install, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 23 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly SystemExit — names the real obstacle, not a sampling gap)

· 21 more finding(s) on lines outside this diff (see the check run).

[safishamsi]

Landed in v0.9.45, just published to PyPI. Cherry-picked onto v8 with your authorship preserved in the commit, so both the change and the credit are in the history. Thanks @ousamabenyounes. Closing since it is now released.