feat(skills): implement federated scientific MCP platform

[kotama7]

Summary

Implements the federated scientific MCP platform designed in #30 and completes the P0–P6 component migration program.

What changed

• Established canonical Skill manifests, immutable run/catalog locks, bounded async lifecycles, explicit node authorization, provider isolation, and a stable five-tool federation surface.
• Integrated pinned ToolUniverse collections plus typed OpenROAD and Qiskit/IBM Quantum profiles without adding tools one by one to the core broker.
• Added scientific-admission, collision/equivalence/independence, EAR/cassette replay, provenance, digest, and fail-closed result contracts.
• Migrated retrieval, measurements, HPC execution, evaluation, memory, plotting/VLM review, paper generation, reproduction/grading, and orchestration to typed contracts.
• Completed P6 removal: deleted retired web aliases/composite collectors, flat measurement writers, low-coverage rubric generation, and container-specific HPC aliases. The scoped slurm_submit core-agent bridge remains under an objective v1.1 removal gate.
• Moved retained-support decisions into permanent English/Japanese/Chinese reference documentation and removed the completed temporary component plan set.
• Fixed documentation link/translation scanners so generated dependencies, VitePress clean URLs, and public assets are handled correctly.

Compatibility and migration

• New producers emit only canonical typed formats.
• Historical readers and narrowly scoped bridges are documented with owners and removal gates in docs/reference/compatibility_support.md.
• Removed producer paths fail explicitly; they are not restored through silent fallbacks.
• Pre-removal rollback tag: skills-p6-pre-removal at c487ea9.

User/developer impact

• Integrators should use search_papers(provider=...), canonical retrieval records, container_submit/job_submit, typed measurement sets, and hierarchical calibrated rubric generation.
• OpenROAD, Qiskit, ToolUniverse, and future MCP collections compose through provider/source adapters and immutable catalog admission rather than per-tool core registration.
• Deprecated tool/config names removed in P6 now return explicit unknown-tool/unknown-field errors.

Validation

• scripts/run_all_tests.sh: 3,752 passed, 0 failed, 30 skipped.
• Frontend: TypeScript check passed; 45 passed, 2 todo.
• Changed Python files: Ruff passed.
• Affected post-cleanup tests: 226 passed, 5 skipped.
• Skill manifests, metadata sync, public/CLI/MCP/viz snapshots, public API, viz API, reference graph, complexity, import boundaries, prompt snapshots, directory policy, and dead-code ratchets passed with no new debt.
• Documentation source, link, source-sync, translation-error, README parity, and VitePress build gates passed.
• git diff --check passed.

Relationship to #30

#30 remains the design RFC. This draft PR is the implementation and migration series built from the separate skills branch, avoiding a non-fast-forward rewrite of the RFC branch.

[github-actions]

Quality Report

• Generated (UTC): 2026-08-03T01:25:16Z
• Repo: /home/runner/work/ARI/ARI
• Checkers: 1 run, 6 unavailable
• Findings: 21 total; 0 net-new vs baseline

Checkers

checker	status	findings	allowlisted	Δ new	note
check_complexity	unavailable	0	0	0	no JSON for check_complexity in quality-artifacts
check_import_boundaries	unavailable	0	0	0	no JSON for check_import_boundaries in quality-artifacts
check_public_api_contracts	error	0	0	0	unparseable JSON (check_public_api_contracts.json): Expectin
check_viz_api_schema	ok	21	21	0
check_prompts	unavailable	0	0	0	no JSON for check_prompts in quality-artifacts
check_dead_code	unavailable	0	0	0	no JSON for check_dead_code in quality-artifacts
check_directory_policy	unavailable	0	0	0	no JSON for check_directory_policy in quality-artifacts

Areas

area	LOC	findings
ari-core/ari/agent	3444	0
ari-core/ari/calibration	17	0
ari-core/ari/cli	2598	0
ari-core/ari/clone	675	0
ari-core/ari/config	973	0
ari-core/ari/configs	69	0
ari-core/ari/evaluator	1365	0
ari-core/ari/llm	1389	0
ari-core/ari/mcp	2475	0
ari-core/ari/memory	353	0
ari-core/ari/migrations	436	0
ari-core/ari/orchestrator	3157	0
ari-core/ari/pipeline	4651	0
ari-core/ari/prompts	420	0
ari-core/ari/protocols	410	0
ari-core/ari/public	834	0
ari-core/ari/publish	898	0
ari-core/ari/registry	511	0
ari-core/ari/schemas	20	0
ari-core/ari/viz	8569	21
ari-skill-benchmark/src	932	0
ari-skill-coding/src	883	0
ari-skill-evaluator/src	825	0
ari-skill-idea/src	2992	0
ari-skill-memory/src	3511	0
ari-skill-orchestrator/src	3642	0
ari-skill-paper/src	5360	0
ari-skill-paper-re/src	8109	0
ari-skill-plot/src	947	0
ari-skill-replicate/src	2636	0
ari-skill-tool-registry/src	12711	0
ari-skill-transform/src	3841	0
ari-skill-vlm/src	976	0
ari-skill-web/src	1510	0

Dead code

• Source: check_dead_code (status: unavailable)
• Note: no JSON for check_dead_code in quality-artifacts
• Baseline: scripts/quality/baselines/dead_code_baseline.json
• Safe-to-delete surviving human review: 0

classification	count
SAFE_DELETE_CANDIDATE	0
QUARANTINE_CANDIDATE	0
TEST_ONLY	0
DOCS_ONLY	0
DYNAMIC_REFERENCE_RISK	0
PUBLIC_CONTRACT	0
REVIEW_REQUIRED	0