feat: add Command Code platform support

[Angelthebestone]

What

Adds Command Code as a first-class supported platform, alongside Claude Code, Codex, Kilo, Trae, Amp, and the rest of the graphify ecosystem.

New platform: command-code

• graphify install --platform command-code — installs the graphify skill to ~/.commandcode/skills/graphify/SKILL.md (user) or .commandcode/skills/graphify/SKILL.md (--project), with the full 8-file references/ progressive-disclosure sidecar.
• graphify command-code install / uninstall — wires the always-on ## graphify section into AGENTS.md (Command Code reads AGENTS.md as project memory) and removes it.
• graphify uninstall — also cleans up the user-global command-code skill.
• Honors COMMANDCODE_CONFIG_DIR as a config-dir override, mirroring CLAUDE_CONFIG_DIR.

How it fits the existing architecture

Command Code discovers skills from .commandcode/skills/ and reads AGENTS.md, so the platform rides the existing agents-md hooks variant with its own caveat-free wording (like amp/agents), the standard Agent-tool dispatch (subagent_type="general") and verbose extraction. The skill is invoked with /graphify exactly like every other platform.

Changes

• skillgen: new [platform.command-code] in platforms.toml, _AGENTS_MD_HOOKS entry, _v8_baseline_ref case (audits against amp's v8 body, like agents), plus regenerated + blessed artifacts (skill-command-code.md, skills/command-code/references/, expected/).
• install.py: _PLATFORM_CONFIG entry, _platform_skill_destination (user/project + env override), project install/uninstall routing, _CLI_INSTALL_COMMANDS, uninstall_all cleanup.
• main.py: help text lists command-code.
• pyproject.toml: description, keyword, and package-data (ships skill-command-code.md).
• tests: platform/destination, wheel payload counts (16 bodies / 112 refs), skillgen progressive host + frontmatter guards.
• README.md: "Pick your platform", "Make your assistant always use the graph", and full command reference tables.
• CHANGELOG.md: entry under 0.9.36 (unreleased).

Validation

• python -m tools.skillgen --check --audit-coverage --schema-singleton --monolith-roundtrip --always-on-roundtrip — clean.
• pytest tests/test_install.py tests/test_install_references.py tests/test_skillgen.py tests/test_install_roundtrip.py tests/test_wheel_packaging.py — 153 passed (platform-related suites).
• Manual end-to-end on Windows: install --platform command-code (user + --project) and graphify command-code install all verified — SKILL.md, 8 references, version stamp, and the AGENTS.md section land correctly.
• Note: 6 unrelated tests fail on Windows only (hermes/gemini LOCALAPPDATA/.agents paths and a codex em-dash encoding assertion) — confirmed pre-existing on a clean checkout; they pass on Linux CI.

[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).

---

Graphify review — findings

This PR adds "Command Code" as a new supported install platform for graphify. It wires up a new --platform command-code option (and graphify command-code install/uninstall subcommands) that copies a new skill-command-code.md skill into ~/.commandcode/skills/graphify/ (user) or .commandcode/skills/graphify/ (project) and manages the ## graphify section in AGENTS.md. Supporting changes touch the install dispatch/uninstall paths, help text, README platform tables, CHANGELOG, and associated skillgen/install tests.

No blocking issues surfaced.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 1066 functions depend on the 893 functions this change touches.

Health — grade B; 10 existing hotspot(s) in the area this change touches (pre-existing, not introduced here):

• dispatch_command() — 2 callers, 118 callees (high)
• codebuddy_install() — 20 callers, 5 callees (high)
• claude_install() — 19 callers, 4 callees (high)
• gemini_install() — 10 callers, 7 callees (high)
• claude_uninstall() — 17 callers, 4 callees (high)
• _project_uninstall() — 5 callers, 13 callees (high)
• render() — 13 callers, 5 callees (high)
• dispatch_install_cli() — 2 callers, 31 callees (high)
• …and 2 more

Verification — 1066 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: 1066 function(s) in the blast radius were not formally verified this run

[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).

---

Graphify review — findings

This PR adds "Command Code" as a new supported install platform for graphify. It introduces a new skill file (skill-command-code.md), wires the platform into the install/uninstall dispatch logic (command-code install/uninstall targeting ~/.commandcode/skills/graphify/ or .commandcode/skills/graphify/ for --project, with AGENTS.md integration), and adds it to the platform config, CLI command set, and help text. Documentation (README, CHANGELOG) and tests are updated to reflect the new platform.

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

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 1078 functions depend on the 895 functions this change touches.

Health — grade B; 10 existing hotspot(s) in the area this change touches (pre-existing, not introduced here):

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

Verification — 1078 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: 1078 function(s) in the blast radius were not formally verified this run