feat(install): add the omp platform for Oh My Pi skill installs

[andrebrait]

What

Adds an omp install platform so graphify's skill can be installed where Oh My Pi (OMP) actually reads agent skills from: ~/.omp/agent/skills/graphify (global) and .omp/agent/skills/graphify (project scope).

Why

OMP mirrors pi's agent layout under ~/.omp, not the legacy ~/.pi path the pi platform writes. graphify install --platform omp therefore failed as an unknown platform, leaving OMP without a graphify skill while every other platform had one. omp was already reserved in the CLI command set as a no-op placeholder; this gives it its implementation.

Changes

• _PLATFORM_CONFIG: new omp entry reusing pi's skill bundle (skill-pi.md + pi's references/ sidecar) — OMP uses the same skill format, so a second bundle would only drift.
• _CLI_INSTALL_COMMANDS: register omp.
• _project_install / _project_uninstall: join the skill-only project group (copilot, pi, kimi, agents).
• dispatch_install_cli: the pi-twin graphify omp [install|uninstall] subcommand.
• tests/test_omp_platform.py: destinations, bundle reuse, end-to-end install/uninstall via the CLI, subcommand behavior, the unknown-platform error listing omp, and uninstall_all parity with pi (the global flag form leaves the user-scope skill to the platform subcommand, exactly as it does for pi).

Test plan

• New suite: 8/8 green (red before the production change).
• Install regression suites: test_agents_platform, test_antigravity_install, test_uninstall_scope, test_wheel_packaging, test_apm_fallback_version, test_atomic_version_stamp — 174 passed.

[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. 1 change(s) tested, no difference found (not proven).

---

Graphify review — findings

Adds an omp (Oh My Pi) install platform that drops the pi skill bundle at ~/.omp/agent/skills/graphify/SKILL.md globally and ./.omp/agent/skills/... in project scope, reusing pi's skill file and references sidecar with no CLAUDE.md. Wires omp into _project_install/_project_uninstall, the CLI install command set, and a graphify omp [install|uninstall] subcommand that mirrors the pi twin (global uninstall goes through the subcommand, not uninstall_all).

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

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 462 functions depend on the 161 functions this change touches.

Health — this change adds coupling hotspots:

• new: main() — 99 callers, 3 callees
• new: dispatch_command() — 2 callers, 125 callees
• new: install() — 13 callers, 11 callees
• new: codebuddy_install() — 20 callers, 5 callees
• new: gemini_install() — 11 callers, 7 callees
• new: claude_install() — 19 callers, 4 callees
• new: _copy_skill_file() — 12 callers, 6 callees
• new: claude_uninstall() — 17 callers, 4 callees
• …and 15 more — each is listed as a finding

Verification — 462 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: 381 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

32 of 301 test file(s) selected (11%) via static blast radius.

• tests/test_affected_cli.py — impact
• tests/test_agents_platform.py — impact
• tests/test_atomic_version_stamp.py — impact
• tests/test_claude_md.py — impact
• tests/test_codebuddy.py — impact
• tests/test_devin.py — impact
• tests/test_explain_cli.py — impact
• tests/test_extract_cli.py — impact
• tests/test_gemini_hook.py — impact
• tests/test_global_add_tag_inference.py — impact
• tests/test_god_nodes_cli.py — impact
• tests/test_hollow_chunks_arm_shrink_guard.py — impact
• tests/test_home_sandbox.py — impact
• tests/test_hook_strict.py — impact
• tests/test_incomplete_build_guard.py — impact
• tests/test_install.py — impact
• tests/test_install_references.py — impact
• tests/test_install_strings.py — impact
• tests/test_install_version_warning.py — impact
• tests/test_merge_chunks_validation.py — impact
• tests/test_multigraph_diagnostics.py — impact
• tests/test_no_dedup_flag.py — impact
• tests/test_omp_platform.py — impact, changed-test
• tests/test_path_cli.py — impact
• tests/test_query_cli.py — impact
• tests/test_query_induced_edges.py — impact
• tests/test_read_hook.py — impact
• tests/test_replace_or_append_section.py — impact
• tests/test_search_hook.py — impact
• tests/test_skill_version_warning.py — impact
• tests/test_uninstall_scope.py — impact
• tests/test_unverified_semantic_shrink.py — impact

Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.

Formal verification

No difference found (not proven): No behavior difference found in dispatch\_install\_cli (not a proof).

The verifier ran both versions of dispatch\_install\_cli on many inputs and saw identical behavior every time. Strong evidence the change is safe, but evidence, not a proof.

Guarantee: Empirical: differential testing (both versions run on many generated inputs). A divergence on an untested input remains possible, so this is 'no counterexample found', not 'proven equivalent'.

Note: An input the sampler did not try could still differ.

Could not verify: Could not verify \_project\_install.

The verifier did not have enough to check \_project\_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: no capturable inputs from the test suite; property tier: not verifiable: all 200 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)

Could not verify: Could not verify \_project\_uninstall.

The verifier did not have enough to check \_project\_uninstall, 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: no capturable inputs from the test suite; property tier: not verifiable: all 200 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly KeyError — names the real obstacle, not a sampling gap)

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

[andrebrait]

Folded into #3506: the omp platform is now part of the OMP native integration PR (the subcommand surface belongs to the plugin installer; skill removal rides uninstall_all). No separate review needed.