merge: sync upstream/main + PR #161 (plugin + compound tools) by d33disc · Pull Request #30 · d33disc/upstream-tooluniverse

d33disc · 2026-04-17T14:39:44Z

Summary

Merges 2 commits from upstream/main (PR Fix/round 155 gaps mims-harvard/ToolUniverse#153 reasoning frameworks + 31 new tools, server.json v1.1.11)
Merges 11 commits from upstream PR Claude Code plugin: self-contained layout, skill-based routing, ML demo readiness mims-harvard/ToolUniverse#161 (Claude Code plugin packaging, compound tools, benchmark harness, skill hardening, torch MPS fix)
All custom code preserved (claude rules, ralph infra, workflows, papers)
Fixed 14 upstream mypy errors (type: ignore comments, defaultdict typing, variable shadowing)

Conflict Resolution

20 conflicts in Phase 1 (squash vs merge of same PR Fix/round 155 gaps mims-harvard/ToolUniverse#153 content) -- resolved by taking upstream for canonical tool files, manual merge for structural files
1 conflict in Phase 2 (.gitignore) -- union of both sides

Verification

2,297 tools loading (up from ~1,200)
All 3 compound tools present (gather_gene_disease_associations, annotate_variant_multi_source, gather_disease_profile)
Plugin directory intact with all components
147 unit tests pass (1 pre-existing failure on main)
Custom code (claude/, ralph-*, paper/) unchanged

Test plan

pytest unit tests pass (147/148, 1 pre-existing failure)
Tool count 2,297
Compound tools load correctly
Plugin directory structure intact
Custom code unchanged

🤖 Generated with Claude Code

…s-harvard#153) Skills (114 total): - Rewrite 80+ skills as reasoning guides (not reference tables) - Add LOOK UP DON'T GUESS and COMPUTE DON'T DESCRIBE across all skills - Add new skills: data-wrangling (24 domain API patterns), dataset-discovery, epidemiological-analysis, data-integration-analysis, ecology-biodiversity, inorganic-physical-chemistry, plant-genomics, vaccine-design, stem-cell, lipidomics, non-coding-RNA, aging-senescence - Add Programmatic Access sections to 6 domain skills (TCGA, GWAS, spatial-transcriptomics, variant-to-mechanism, binder-discovery, clinical-trials) - Generalize all analysis skills to be data-source-agnostic - Add progressive disclosure: references/ for specialized domains - Improve skill descriptions for better triggering Tools (31 new): - RGD (4 tools), T3DB toxins, IEDB MHC binding prediction - 11 scientific calculator tools (DNA translate, molecular formula, equilibrium solver, enzyme kinetics, statistics, etc.) - AgingCohort_search (28+ longitudinal cohort registry) - NHANES_download_and_parse (XPT download + parse + age filter) - DataQuality_assess (missingness, outliers, correlations) - MetaAnalysis_run (fixed/random effects, I-squared, Q-test) - 4 dataset discovery tools (re3data, Data.gov, OpenAIRE, DataCite) Bug fixes: - Fix 50+ tool name references across skills - Fix NHANES search (dynamic CDC catalog query, not hardcoded keywords) - Fix tool return envelopes (Unpaywall, MyGene, HPA, EuropePMC) - Fix STRING, OpenTargets, ENCODE, Foldseek, STITCH, BridgeDb - Fix BindingDB test for broken API detection Router: - Add MC elimination strategy, batch processing protocol - Add 20+ bundled computation scripts - Route to all 114 skills Version bumped to 1.1.11

New plugin/ directory with official Claude Code plugin format: - .claude-plugin/plugin.json: manifest (name, version, description) - .mcp.json: auto-configures ToolUniverse MCP server with --refresh - settings.json: auto-approve read-only discovery tools - commands/find-tools.md: /tooluniverse:find-tools slash command - commands/run-tool.md: /tooluniverse:run-tool slash command - agents/researcher.md: autonomous research agent with 1000+ tools - README.md: install and usage documentation Build script: scripts/build-plugin.sh - Assembles distributable plugin from repo (manifest + skills + agents) - Copies all 113 tooluniverse-* skills into plugin/skills/ - Output: dist/tooluniverse-plugin/ (7.6MB, 520 files) Install: claude --plugin-dir dist/tooluniverse-plugin

gene-regulatory-networks and population-genetics had markdown headings instead of YAML frontmatter, preventing Claude Code skill discovery.

Addressed 4 weaknesses found in A/B testing: 1. Reduce discovery overhead: Added example parameters to all tools in quick reference — agent can call directly without get_tool_info 2. Enforce batching: Added explicit Python batch pattern with code example in both research command and researcher agent 3. Prevent trial-and-error: Added exact parameter formats (e.g., OncoKB needs "operation" field, OpenTargets needs ensemblId not gene symbol) 4. Added /tooluniverse:research command — comprehensive slash command with full tool reference table and efficiency rules Test results: find_tools calls reduced 75% (4→1), subagent spawns eliminated, cross-validation now happening across 4 databases.

MCP is good for tool discovery (find_tools, get_tool_info) but inefficient for batch data retrieval (37 sequential execute_tool calls). Changed strategy: use CLI (tu run) via Python scripts for all actual data retrieval. One Python script with 10 tu_run() calls replaces 10 sequential MCP calls. MCP reserved for discovery only. Updated: researcher agent, research command, find-tools command, README. Added tu_run() helper function pattern and Python SDK example.

…ketplace - plugin/skills/ now contains per-skill symlinks to ../../skills/tooluniverse-* + setup-tooluniverse so the plugin directory is self-contained without moving the source skills/ folder. - plugin/sync-skills.sh regenerates the symlink set when skills are added. - plugin/.claude-plugin/marketplace.json declares the plugin dir as a single-plugin marketplace, enabling 'claude plugin install tooluniverse@tooluniverse-local' workflow. - .gitignore excludes benchmark outputs (skills/evals/*/results_*.json), memory notes, and API-key patterns from the repo. - .gitattributes adds export-ignore for non-plugin directories so 'git archive' produces a clean release tarball.

… content commands/research.md is now scoped to TU usage (tool recipes, compound tools, skill dispatch table). Domain analysis guidance moved into the matching specialized skills so content has a single owner. Skill additions (each skill gains a 'BixBench-verified conventions' section): - tooluniverse-statistical-modeling: clinical-trial AE inner-join pattern, OR reduction semantics, F-stat vs p-value distinction, spline pure-strain anchor, frequency-ratio output format, CSV latin1 fallback. - tooluniverse-rnaseq-deseq2: authoritative-script pattern (copy ALL kwargs literally incl. refit_cooks=True), R vs pydeseq2 selection rule, strain identity parsing, 'uniquely DE' exclusive semantics, denominator check for set-operation percentages. - tooluniverse-gene-enrichment: R clusterProfiler vs gseapy selection, simplify(0.7) term-collapse caveat, explicit universe= background rule. - tooluniverse-crispr-screen-analysis: sgRNA-level Spearman convention, Reactome GSEA ranking column, literal pathway-name matching. - tooluniverse-phylogenetics: parsimony informative site gap-only exclusion, treeness ratio definition. - tooluniverse-variant-analysis: multi-row Excel header parsing, SO-term coding vs non-coding denominator split. tooluniverse-drug-target-validation improvements for the ML demo: - Top-level 'RUN THE ML MODELS, DON'T SKIP THEM' rule alongside 'LOOK UP DON'T GUESS'. - New Phase 3b requiring all 10 ADMET-AI Chemprop-GNN endpoints and a side-by-side head-to-head table when multiple candidate compounds exist. - Phase 8 now mandates ESMFold + DoGSite3 (ProteinsPlus) even when PDB structures exist, so the deep-learning inference is always in the trace. - Phase 10 adds a 'Deep-Learning Models Contributing' attribution table naming each ML predictor's architecture and contribution.

ADMET-AI tools segfaulted (exit 139) via tu CLI / MCP server on macOS Apple Silicon. Root cause: torch MPS backend crashes in forked subprocess. Fix: torch.set_default_device('cpu') at package init + env vars.

research.md: add skill dispatch table at top so /tooluniverse:research routes cancer-mutation queries to precision-oncology, target-validation queries to drug-target-validation, etc. precision-oncology: promote FAERS to MANDATORY (was optional bullet). Agent now calls FAERS_search_adverse_event_reports for top 1-2 drugs before finalizing. drug-target-validation: add ADMET-AI SDK fallback pattern — if MCP calls fail, agent retries via Python SDK in Bash. .mcp.json: add PYTORCH env vars for MPS fallback.

Make Claude Code plugin installation a two-command flow: claude plugin marketplace add mims-harvard/ToolUniverse claude plugin install tooluniverse@tooluniverse Changes: - .claude-plugin/marketplace.json at repo root with source: ./plugin (enables GitHub owner/repo marketplace add without sparse checkout) - skills/tooluniverse-install-plugin/SKILL.md: user-facing install guide (prereqs, two-command install, version pinning, verify, API keys, update/uninstall, offline zip path, troubleshooting table) - .github/workflows/release-plugin.yml: on tag push, build tooluniverse-plugin-vX.Y.Z.zip with resolved skills symlinks and a rewritten marketplace.json, attach to the GitHub release - plugin/README.md: replace local path install with marketplace flow, link to the install skill - skills/setup-tooluniverse/SKILL.md: callout for Claude Code users pointing at the plugin install path over manual MCP config

The install skill is Claude-Code-plugin-specific, so name it that way — `tooluniverse-install-plugin` was ambiguous (install what? which plugin?). Renamed directory + frontmatter name + all inbound refs in plugin/README.md, setup-tooluniverse skill, and the release workflow.

Implements the plan for improving plugin output quality on multi- database questions: Compound tools (3 new, each aggregates multiple atomic databases): - gather_gene_disease_associations — DisGeNET + OMIM + OpenTargets + GenCC + ClinVar with cross-source concordance scoring - annotate_variant_multi_source — ClinVar + gnomAD + CIViC + UniProt - gather_disease_profile — Orphanet + OMIM + DisGeNET + OpenTargets + OLS, returns unified identifiers (orphanet/omim/efo/mondo) + gene associations These return structured {status, data} with a sources_failed list, so partial failures are tolerated without the whole call erroring. MSigDB tool + config: - check_gene_in_set / get_gene_set_members operations covering GTRD TF targets, miRDB miRNA targets, oncogenic sigs (C6), hallmarks (H) Benchmark harness skill (skills/devtu-benchmark-harness): - run_eval.py — unified runner for lab-bench + BixBench, with --mode, --category, --n, --timeout; resumes from existing results - grade_answers.py — exact / MC / range / normalized / numeric / LLM-verifier strategies, batch grading - analyze_results.py — category accuracy, per-q plugin-vs-baseline delta, failure classification (timeout / error / wrong / grading) - generate_report.py — markdown report with exec summary + top failures - Phase 3.5 in devtu-self-evolve invokes the harness after testing Plumbing: - _lazy_registry_static.py: 4 new tool class entries - default_config.py: 3 new JSON paths for compound tools - skills/evals: question banks for bixbench (61 Q) and lab-bench (20 Q) checked in; result snapshots gitignored - tests/test_claude_code_plugin.py: 700 lines validating plugin manifest / MCP / settings / commands / agent / tool refs - tests/test_aging_cohort_tool.py: 385 lines for AgingCohort tool

…1.11) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

…ompound tools) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Merges upstream changes and adds/updates ToolUniverse plugin packaging, compound tools, and benchmark harness assets while preserving fork customizations.

Changes:

Adds Claude Code plugin packaging (plugin manifests/config, build/release workflow) and updates skills/docs for plugin installation and benchmark conventions.
Introduces compound tools for multi-source gene/disease/variant queries plus new MSigDB and NHANES functionality, and registers new tools in configs/registries.
Updates numerous tool JSON schemas/examples and adds benchmark harness scripts + eval artifacts.

Reviewed changes

Copilot reviewed 191 out of 193 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
tests/tools/test_semantic_scholar_tool_resilience.py	Removes Semantic Scholar rate-limit regression tests; keeps basic error-shaping/abstract enrichment tests.
tests/test_aging_cohort_tool.py	Adds unit tests for AgingCohort search behavior and cohort registry completeness.
src/tooluniverse/tools/init.py	Registers additional tool wrappers (BioGRID tools, NHANES_download_and_parse).
src/tooluniverse/tools/NHANES_download_and_parse.py	Refactors wrapper arg-building and signature for NHANES download/parse.
src/tooluniverse/tools/MetaAnalysis_run.py	Tightens typing/docs and wrapper signature for meta-analysis tool.
src/tooluniverse/tools/DegreesOfUnsaturation_calculate.py	Adjusts parameter ordering/docs and request arg mapping for DoU tool wrapper.
src/tooluniverse/tools/DataQuality_assess.py	Tightens typing/docs and wrapper signature for data quality tool.
src/tooluniverse/tools/AgingCohort_search.py	Improves typing and docstring for aging cohort search wrapper.
src/tooluniverse/restful_tool.py	Removes response-size trimming logic for closure arrays.
src/tooluniverse/msigdb_tool.py	Adds a new MSigDB tool implementation (gene set fetch/search/membership).
src/tooluniverse/default_config.py	Adds compound tool JSON files to default tool file set.
src/tooluniverse/data/zfin_tools.json	Adds/updates ZFIN search tool schema/examples (now duplicated).
src/tooluniverse/data/wikipathways_tools.json	Adjusts test examples for WikiPathways tool.
src/tooluniverse/data/unpaywall_tools.json	Updates Unpaywall test examples.
src/tooluniverse/data/rcsb_advanced_search_tools.json	Refactors return schema to oneOf success/error forms.
src/tooluniverse/data/nhanes_tools.json	Adds NHANES_download_and_parse tool definition.
src/tooluniverse/data/msigdb_tools.json	Adds MSigDB tool entries for membership checks + gene set retrieval.
src/tooluniverse/data/mgi_tools.json	Refactors MGI schemas to oneOf success/error + metadata wrapping.
src/tooluniverse/data/iedb_tools.json	Removes antigen_uniprot parameter/mapping from IEDB tool schema.
src/tooluniverse/data/europe_pmc_tools.json	Adds structured full-text retrieval tool schema for Europe PMC.
src/tooluniverse/data/ena_portal_tools.json	Refactors ENA portal schemas to oneOf success/error + metadata wrapping.
src/tooluniverse/data/datacite_tools.json	Adjusts DataCite schema fields and changes search return schema shape.
src/tooluniverse/data/compound_variant_tools.json	Adds compound variant annotation tool schema.
src/tooluniverse/data/compound_gene_disease_tools.json	Adds compound gene–disease association tool schema.
src/tooluniverse/data/compound_disease_tools.json	Adds compound disease profile tool schema.
src/tooluniverse/data/brenda_tools.json	Adds/updates BRENDA enzyme kinetics tool schema (now duplicated).
src/tooluniverse/data/admetai_tools.json	Reduces/edits ADMET-AI test examples.
src/tooluniverse/compound_variant_tool.py	Implements compound variant annotation tool.
src/tooluniverse/compound_gene_disease_tool.py	Implements compound gene–disease association tool.
src/tooluniverse/compound_disease_tool.py	Implements compound disease profile tool.
src/tooluniverse/admetai_tool.py	Adds torch/MPS safeguards and type-ignore annotations around DataFrame usage.
src/tooluniverse/_lazy_registry_static.py	Registers new tool classes for lazy loading (compound tools, MSigDB, DataQualityTool).
src/tooluniverse/init.py	Adds global torch/MPS environment + default-device forcing at import time.
skills/tooluniverse-variant-analysis/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-statistical-modeling/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-rnaseq-deseq2/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-precision-oncology/SKILL.md	Strengthens mandatory safety/pharmacogenomics steps.
skills/tooluniverse-population-genetics/SKILL.md	Adds YAML frontmatter metadata for plugin consumption.
skills/tooluniverse-phylogenetics/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-gene-regulatory-networks/SKILL.md	Adds YAML frontmatter metadata for plugin consumption.
skills/tooluniverse-gene-enrichment/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-drug-target-validation/SKILL.md	Adds mandatory ML-model execution guidance and reporting requirements.
skills/tooluniverse-crispr-screen-analysis/SKILL.md	Adds BixBench-verified conventions section.
skills/tooluniverse-claude-code-plugin/SKILL.md	Adds new skill documenting plugin installation/maintenance.
skills/setup-tooluniverse/SKILL.md	Updates setup guidance to prefer plugin install for Claude Code users.
skills/evals/run_benchmark.py	Adds benchmark runner for Claude Code plugin evaluation.
skills/evals/research_eval_results.json	Adds evaluation results artifact.
skills/evals/lab-bench/questions.json	Adds lab-bench question set.
skills/devtu-self-evolve/SKILL.md	Adds benchmark evaluation phase guidance.
skills/devtu-benchmark-harness/scripts/run_eval.py	Adds unified benchmark runner script.
skills/devtu-benchmark-harness/scripts/grade_answers.py	Adds multi-strategy answer grading script.
skills/devtu-benchmark-harness/scripts/generate_report.py	Adds markdown report generator.
skills/devtu-benchmark-harness/scripts/analyze_results.py	Adds results analyzer and failure categorization.
skills/devtu-benchmark-harness/references/benchmark-guide.md	Adds benchmark guide/reference document.
skills/devtu-benchmark-harness/evals/evals.json	Adds eval definitions for harness.
skills/devtu-benchmark-harness/SKILL.md	Adds new skill documenting benchmark harness workflow.
scripts/build-plugin.sh	Adds local build script for assembling plugin dist directory.
plugin/sync-skills.sh	Adds script to symlink user-facing skills into plugin dir.
plugin/settings.json	Adds default MCP permission auto-approve settings.
plugin/commands/run-tool.md	Adds slash command for executing a named tool.
plugin/commands/research.md	Adds primary research slash command guidance and routing rules.
plugin/commands/find-tools.md	Adds tool discovery slash command.
plugin/agents/researcher.md	Adds autonomous researcher agent definition.
plugin/README.md	Adds plugin README and usage notes.
plugin/.mcp.json	Adds MCP server config for uvx tooluniverse with env vars.
plugin/.claude-plugin/plugin.json	Adds plugin manifest.
plugin/.claude-plugin/marketplace.json	Adds local marketplace manifest for plugin.
.gitignore	Ignores benchmark outputs and local memory/secrets.
.github/workflows/release-plugin.yml	Adds GitHub Action to build and attach plugin zip on tags.
.gitattributes	Adds export-ignore rules for archives/releases.
.claude-plugin/marketplace.json	Adds root marketplace manifest pointing to ./plugin source.

Comments suppressed due to low confidence (2)

src/tooluniverse/data/brenda_tools.json:304

This file now defines BRENDA_get_enzyme_kinetics twice (second copy begins around line 296). Duplicate tool names can cause registry collisions and unpredictable loading. Remove the duplicate entry or consolidate the definitions into one.
src/tooluniverse/data/zfin_tools.json:666
This file now contains two separate tool objects with the same name ZFIN_search (around lines 603 and 661). Duplicate tool names can lead to ambiguous/last-one-wins behavior when loading the registry. Remove one of the duplicate entries (or rename if they are intended to differ).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-17T14:45:18Z

+        _args["variables"] = variables
+    if age_min is not None:
+        _args["age_min"] = age_min
+    if age_max is not None:
+        _args["age_max"] = age_max
+    if max_rows != 5000:
+        _args["max_rows"] = max_rows


max_rows is annotated as int, but callers can still pass JSON null which becomes None. In that case, max_rows != 5000 is true and _args['max_rows'] will be set to None, likely failing schema validation downstream. Consider either restoring Optional[int] and guarding with if max_rows is not None and max_rows != 5000, or coercing/validating max_rows before building _args.

Copilot · 2026-04-17T14:45:28Z

+# Force CPU before torch is imported anywhere — prevents MPS (Metal) segfaults
+# in forked subprocesses (uvx MCP server, tu CLI, Claude Code plugin).
+os.environ.setdefault("PYTORCH_MPS_HIGH_WATERMARK_RATIO", "0.0")
+os.environ.setdefault("PYTORCH_ENABLE_MPS_FALLBACK", "1")
+try:
+    import torch
+
+    if hasattr(torch, "set_default_device"):
+        torch.set_default_device("cpu")
+except ImportError:


Importing torch at package import time adds a heavy dependency load and global side effects (setting default device) even for users who never invoke ADMET-AI. This can slow ToolUniverse startup and can prevent GPU/MPS usage in other contexts. Consider limiting the package-level change to setting the env vars, and defer any torch import / set_default_device('cpu') to the ADMET-related code paths only (e.g., inside admetai_tool.py).

Copilot · 2026-04-17T14:45:28Z

+        from .execute_function import ToolUniverse
+
+        tu = ToolUniverse()
+        tu.load_tools()
+


Creating a new ToolUniverse() and calling load_tools() inside every run() will reload the full tool registry per invocation (potentially thousands of tools), which is expensive and can become a major latency bottleneck for this compound tool. Consider reusing a shared/singleton ToolUniverse instance (module-level cache) or calling the underlying tool registry/client directly to avoid repeated full loads.

Copilot · 2026-04-17T14:45:29Z

        response = remove_none_and_empty_values(response)
-
-        # Strip ontology closure arrays — they contain ~150 ancestor terms per
-        # item and inflate responses from ~5KB to 100KB+ without adding signal
-        # for downstream LLM consumption.
-        _CLOSURE_KEYS = {"object_closure", "object_closure_label"}
-
-        def _strip_closure(obj):
-            if isinstance(obj, dict):
-                return {
-                    k: _strip_closure(v)
-                    for k, v in obj.items()
-                    if k not in _CLOSURE_KEYS
-                }
-            if isinstance(obj, list):
-                return [_strip_closure(v) for v in obj]
-            return obj
-
-        response = _strip_closure(response)
        if isinstance(response, dict) and "status" not in response:
            return {"status": "success", "data": response}
        return response


The response post-processing no longer strips object_closure / object_closure_label arrays. Those fields can be very large and were previously removed to keep responses small for downstream consumption. If this was removed unintentionally during the upstream merge, consider restoring the stripping (optionally behind a flag) to avoid large payloads and higher latency/memory usage.

Copilot · 2026-04-17T14:45:29Z

+for skill_dir in "$REPO_ROOT/skills"/*/; do
+    dir_name=$(basename "$skill_dir")
+    if [ -f "$skill_dir/SKILL.md" ] && [[ "$dir_name" == tooluniverse* ]]; then
+        cp -r "$skill_dir" "$DIST_DIR/skills/$dir_name"
+        skill_count=$((skill_count + 1))


The local plugin build script only copies skills whose directory name starts with tooluniverse*, but plugin/sync-skills.sh includes setup-tooluniverse. This means a locally-built plugin will be missing the setup skill (and any others outside the prefix). Update the filter to include setup-tooluniverse (and any other intended user-facing skills) so local builds match the release artifact behavior.

Copilot · 2026-04-17T14:45:29Z

+        from .execute_function import ToolUniverse
+
+        tu = ToolUniverse()
+        tu.load_tools()
+


Creating a new ToolUniverse() and calling load_tools() on every run() reloads the full tool registry per call and will be a significant latency hit for this compound tool. Consider caching a ToolUniverse instance or using a lighter-weight execution path to call the underlying tools without re-loading everything each time.

Copilot · 2026-04-17T14:45:30Z

+        from .execute_function import ToolUniverse
+
+        tu = ToolUniverse()
+        tu.load_tools()
+


Creating a new ToolUniverse() and calling load_tools() on every run() reloads the full tool registry per call and will be a significant latency hit for this compound tool. Consider caching a ToolUniverse instance or using a lighter-weight execution path to call the underlying tools without re-loading everything each time.

Copilot · 2026-04-17T14:45:30Z

    _args = {
        k: v
        for k, v in {
            "operation": operation,
            "formula": formula,
            "C": C,
            "H": H,
            "N": N,
+            "O": oxygen,
            "S": S,
            "F": F,
            "Cl": Cl,
            "Br": Br,
-            "oxygen": oxygen,
-            "iodine": iodine,
+            "I": iodine,
        }.items()


The request payload keys no longer match the tool schema: degrees_of_unsaturation_tools.json defines parameters oxygen and iodine, but this wrapper sends them as O and I. This will cause the shared client to reject/ignore oxygen/iodine arguments. Align the argument dict keys with the JSON tool schema (or update the schema to match).

gasvn and others added 15 commits March 29, 2026 18:41

chore: sync server.json version to 1.1.11 [skip ci]

4d66869

fix: add missing YAML frontmatter to 2 skills

3c58148

gene-regulatory-networks and population-genetics had markdown headings instead of YAML frontmatter, preventing Claude Code skill discovery.

fix: force torch CPU to prevent MPS segfault in subprocess

e760cfa

ADMET-AI tools segfaulted (exit 139) via tu CLI / MCP server on macOS Apple Silicon. Root cause: torch MPS backend crashes in forked subprocess. Fix: torch.set_default_device('cpu') at package init + env vars.

merge: sync with upstream/main (PR mims-harvard#153 + server.json v1.…

2c99ee4

…1.11) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

merge: integrate upstream PR mims-harvard#161 (Claude Code plugin + c…

428917d

…ompound tools) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Copilot AI review requested due to automatic review settings April 17, 2026 14:39

github-actions Bot added documentation Improvements or additions to documentation core tools tests github-actions labels Apr 17, 2026

Copilot started reviewing on behalf of d33disc April 17, 2026 14:40 View session

Copilot AI reviewed Apr 17, 2026

View reviewed changes

d33disc merged commit c5df4c2 into main Apr 17, 2026
6 checks passed

d33disc deleted the feat/upstream-sync-2026-04-17 branch April 17, 2026 14:53

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

merge: sync upstream/main + PR #161 (plugin + compound tools)#30

merge: sync upstream/main + PR #161 (plugin + compound tools)#30
d33disc merged 15 commits intomainfrom
feat/upstream-sync-2026-04-17

d33disc commented Apr 17, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Copilot AI Apr 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

d33disc commented Apr 17, 2026

Summary

Conflict Resolution

Verification

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants