Verification updates

Rolling, reverse-chronological log of catalog verification + security passes. Each scheduled or on-demand run that produces changes prepends a dated block; the top block is mirrored to the pinned “Verification updates” issue.

Older entries live in VERIFIER_CHANGELOG_ARCHIVE.md.

2026-08-17

Verified

  • Adjudicated the injected 15-page digest: acmg-classification, alignment-trimming, alkyl, blatant-why, can-immune, mcptools, medicare-mcp, msa-statistics, msi-detection, multiple-alignment, neuro-mcp, nwb-mcp-server, pbmcpedia, somatic-signatures, structural-alignment. All 15 graded works; 13 caution, 2 cleared (mcptools, medicare-mcp). 9 further digest pages were over this run’s 15-page review budget and were intentionally left untouched.
  • Confirmed install/launch commands against primary sources for every non-smoke-tested page in the batch: mcptools (package vignette), medicare-mcp (package.json + README), pbmcpedia (server.ts + README) all match the catalog pages verbatim.

Fixed

  • No install-path or launch-command defects found this run; no page-content fixes were required beyond the stamp fields and table rows.

Flagged

  • Discrepancy vs. prior runs: GPTomics/bioSkills is now confirmed archived: true on GitHub (1187★, pushed 2026-08-15). This reverses the not-archived status this same repo has carried across 5+ prior verifier runs (most recently 2026-08-13). All 7 affected catalog pages (acmg-classification, alignment-trimming, msa-statistics, msi-detection, multiple-alignment, somatic-signatures, structural-alignment) are graded works/caution this run — MIT root license and skill directories are still intact and resolve, but no further upstream maintenance should be expected. Flagged in catalog/verifier-state.md for next-run recheck.
  • 9 digest pages remain over this run’s review budget and stay due for the next run.

Security

  • Resolved 3 digest flags as non-issues on primary-source evidence: osv-advisory on requests (acmg-classification) and on uv (nwb-mcp-server) both trace to GHSA fixes that shipped before each page’s pinned version; license-unrecognized (mcptools) resolved via a raw LICENSE.md fetch confirming standard MIT (GitHub’s classifier had simply failed to auto-detect it) — regraded cleared.
  • endpoint-non-2xx on can-immune confirmed as expected live-server behavior (406 to a browser-shaped GET, consistent with the digest’s own 405 finding).
  • New caution grades (first review, provenance/license clean but carrying an independent risk signal): alkyl (single-maintainer Beta, no CI-passing tests), blatant-why (optional external compute/wet-lab API keys), can-immune (new/low-traffic repo over COSMIC/DepMap-derived data), neuro-mcp (Alpha/0-star single-org project persisting subject and EHR records locally), nwb-mcp-server (single-maintainer, pins a pre-release dependency), pbmcpedia (stale, low-traffic repo).
  • New cleared grades: mcptools (MIT confirmed), medicare-mcp (MIT confirmed, provenance and launch command verified against package.json/README).

2026-08-13

Verified

  • Stamped 15 previously-unreviewed catalog pages with full verification/security badges: loop-calling, metabolite-communication, multiplicity-graphical, ncrna-search, netneurotools-guide, neural-population-analysis-guide, parameter-recovery-checker, power-and-sample-size, pycortex-guide, structure-probing, subgroup-analysis, tad-detection, tooluniverse-clinical-trial-design, tooluniverse-clinical-trial-matching, tooluniverse-drug-mechanism-research — all graded works/cleared.
  • 8 GPTomics/bioSkills pages confirmed against a fresh root LICENSE fetch this run (MIT verbatim, 1.2k★, not archived): loop-calling, metabolite-communication, multiplicity-graphical, ncrna-search, power-and-sample-size, structure-probing, subgroup-analysis, tad-detection. None depend on external credentials.
  • 4 HaoxuanLiTHUAI/awesome_cognitive_and_neuroscience_skills pages (netneurotools-guide, neural-population-analysis-guide, parameter-recovery-checker, pycortex-guide) carried a repo-renamed flag; confirmed this run as the same genuine GitHub org transfer to NeuroAIHub already established for sibling skills in this collection (old owner URL still resolves and redirects to the new owner) — no page fix needed.
  • 3 ToolUniverse skill pages (tooluniverse-clinical-trial-design, tooluniverse-clinical-trial-matching, tooluniverse-drug-mechanism-research) confirmed against a fresh mims-harvard/ToolUniverse fetch this run (Apache-2.0, active, 1,624★); each is read-only over public sources (Open Targets, ClinicalTrials.gov, ChEMBL, KEGG, Reactome, STRING, DailyMed, PharmGKB, EU CTIS, ISRCTN, CIViC, MyGene, FDA/openFDA/FAERS, PubMed/Europe PMC) with no API-key/credential dependency, matching the pattern of the ~22 already-stamped ToolUniverse pages.

Fixed

  • None — all 15 pages resolved cleanly with no install-path, launch-command, or licence discrepancies.

Flagged

  • None new. Two additional pages in this run’s digest were over the 15-page review budget and were intentionally left unstamped; they lead the next worklist.

Security

  • All 15 pages graded cleared: provenance matches supplier for every page, real permissive licenses (MIT ×12, Apache-2.0 ×3) confirmed by fresh fetch, no OSV/GHSA advisories surfaced, and no external-service credential requirements found in any of the three ToolUniverse skills reviewed.

2026-08-10

Verified

  • Worked the injected 15-page adjudication digest at this run’s 15-page review budget: bayesian-trials, cdisc-data-handling, chemgraph, clustering-phenotyping, compartment-analysis, compensation-transformation, covariation-analysis, cytometry-differential-analysis, cytometry-qc, effect-measures, fda-mcp, gating-analysis, geometric-analysis, hashing-demultiplexing, lesion-symptom-mapping-guide. All first-time full stamps (never previously reviewed); all graded works.
  • 9 GPTomics/bioSkills pages confirmed against a fresh root LICENSE fetch (MIT verbatim) → works/cleared.
  • chemgraph (argonne-lcf) — Apache-2.0 confirmed via raw LICENSE fetch, provenance matches ALCF, python -m chemgraph.mcp.mcp_tools launch command confirmed verbatim in upstream README, no OSV/GHSA advisories.
  • fda-mcp (openpharma-org) — MIT confirmed via raw LICENSE fetch, provenance matches (OpenPharma confirmed as a real 50+-repo coordinated org), node build/index.js stdio launch already correctly documented on the page.
  • lesion-symptom-mapping-guide — repo-renamed flag confirmed a genuine GitHub org transfer HaoxuanLiTHUAI → NeuroAIHub, same pattern already established for sibling skills in this collection.
  • covariation-analysis — R-scape dependency confirmed GPLv3 (EddyRivasLab) this run.
  • geometric-analysis — DSSP dependency confirmed migrated to Boost Software License / BSD-2-Clause (no longer restrictive) this run.

Fixed

  • No page-content fixes were needed this run — all evidence gathered (GPTomics MIT root, chemgraph Apache-2.0, fda-mcp MIT, lesion-symptom-mapping-guide repo-renamed) confirmed the pages were already accurate.

Flagged

  • cdisc-data-handling added to the Flagged list: security caution — skill code MIT/clean and provenance confirmed, but CDISC standards require membership/licence for some deliverables and Pinnacle 21 Enterprise is commercial (free Community validator is more limited). Data/tool-license restriction only, same pattern as kegg-pathway-analysis. Verification works.
  • 17 further flagged pages were over this run’s review budget and were intentionally left untouched — they should lead the next worklist.

Security

  • 14 of 15 pages graded cleared; 1 graded caution (cdisc-data-handling, data-use restriction only — see Flagged).
  • No new advisories found via WebSearch/GitHub security-advisories for GPTomics/bioSkills, argonne-lcf/ChemGraph, or openpharma-org/fda-mcp.

2026-08-06

Verified

  • Worked the injected 15-page adjudication digest: omophub-mcp, optogenetics-protocol-designer, perturb-seq, pyomop, rosetta-mcp-server, scatac-analysis, signal-detection-analysis, strain-tracking, structure-preparation, structure-validation, tooluniverse-admet-prediction, tooluniverse-cell-line-profiling, tooluniverse-chemical-sourcing, tooluniverse-dose-response, trial-reporting.
  • perturb-seq, scatac-analysis, strain-tracking, trial-reporting (GPTomics bioSkills) — works/cleared, fresh root LICENSE fetch confirms MIT.
  • tooluniverse-admet-prediction, tooluniverse-cell-line-profiling, tooluniverse-chemical-sourcing, tooluniverse-dose-response (ToolUniverse) — works/cleared, confirmed against mims-harvard/ToolUniverse Apache-2.0.
  • optogenetics-protocol-designer, signal-detection-analysis (awesome_cognitive_and_neuroscience_skills) — works/cleared; the repo-renamed flag on both confirmed as a legitimate GitHub org transfer to NeuroAIHub, not a typosquat.
  • rosetta-mcp-server — works/caution; install/launch commands confirmed unchanged against the README.

Fixed

  • pyomop — the documented pyomop-mcp-server console-script entry point boot-errored in this run’s smoke test (No such file or directory) despite a clean pip install pyomop. Replaced the install/registration blocks (Verify-it-starts, Claude Code stdio, Claude Desktop config) with the working pyomop --mcp-server subcommand form. Graded degraded pending a clean reboot confirmation next run.

Flagged

  • structure-preparation, structure-validation (GPTomics bioSkills) — security caution: bundled Phenix/CCTBX tooling (reduce, phenix.molprobity, phenix.process_predicted_model) is free for academic use only, same pattern as mixcr-analysis.
  • rosetta-mcp-server — security caution: license-absent flag confirmed accurate, no LICENSE file committed upstream, MIT asserted only in README/package.json prose.
  • omophub-mcp — verification degraded: auth-gated (requires a signed-up OMOPHUB_API_KEY), so functionally unverifiable without an account.
  • pyomop — verification degraded pending a clean-reboot recheck of pyomop --mcp-server next run.
  • One further flagged page from this run’s digest was over the 15-page review budget and was intentionally left untouched; it leads the next worklist.

Security

  • pyomop — license-absent flag resolved: fetched the repository LICENSE directly from the develop branch (the actual GitHub default branch) and confirmed GPL-3.0 verbatim, matching the GitHub API license field. Regraded security cleared.
  • signal-detection-analysis — osv-advisory flag traced to three old scipy GHSA IDs, all fixed-by-or-withdrawn before the pinned scipy 1.18.0; no live advisory applies. Regraded security cleared.

2026-08-03

Verified

  • First-time full review + stamp of a 15-page bootstrap batch: adaptive-designs, binding-site-detection, calcium-imaging-analysis-guide, cnv-inference, deeplabcut, differential-abundance, doublet-detection, drift-diffusion-model, functional-profiling, immunogenicity-scoring, interface-analysis, lineage-tracing, materials-project-mcp, metaphlan-profiling, missing-data-sensitivity. All graded works.
  • GPTomics/bioSkills (10 pages) reconfirmed MIT via a fresh root LICENSE fetch (1.1k★, not archived).
  • HaoxuanLiTHUAI/awesome_cognitive_and_neuroscience_skills (3 pages) repo-transfer to NeuroAIHub confirmed genuine (old URL redirects live, new README still documents old install path) — no fix needed.
  • materials-project-mcp launch command uvx mcp-science materials-project reconfirmed against PyPI mcp-science 0.1.7 metadata and the mcp.science root README (MIT, 146★, no advisories).

Fixed

  • None — all three inspection flags this run (repo-renamed ×3, smoke-install-error ×1) were resolved as non-issues on evidence; no page content required correction.

Flagged

  • immunogenicity-scoring (GPTomics bioSkills) — security caution: skill code MIT/clean, but bundled MixMHCpred/PRIME predictors are academic/non-commercial-use only (same pattern as mixcr-analysis).

Security

  • lineage-tracing — smoke job’s install_error is a false negative: it pip-installed the stale/deprecated PyPI cassiopeia-lineage 1.0.4 (2019, incompatible with Python 3.12’s removed numpy.distutils), a path the catalog page itself already warns against. The page’s own recommended pip install git+https://github.com/YosefLab/Cassiopeia@master was confirmed live. Graded works/cleared.
  • No new advisories found for any of the 15 pages’ upstream repos/packages (GitHub /security pages and PyPI metadata checked where applicable).

2026-07-29

Verified

  • arboreto (K-Dense) — works/cleared, verified_on/reviewed_on refreshed to 2026-07-29. Liveness flagged changed-since-verified (skill dir last committed 2026-07-26): fetched skills/arboreto/SKILL.md this run — unchanged install (uv pip install arboreto / conda install -c bioconda arboreto), no risky patterns, no external creds. Smoke-test pip install arboreto passed this run (.verify/smoke-results.json), upgrading the evidence to a real install verdict.
  • archs4-database (SciAgent) — works/cleared, verified_on/reviewed_on refreshed to 2026-07-29. Liveness flagged license-missing: GitHub’s API reports NOASSERTION for jaechang-hits/SciAgent-Skills, but the raw root LICENSE fetched this run is verbatim CC BY 4.0 — consistent with the standing SciAgent finding, reconfirmed independently this run.
  • arxiv — works/cleared, verified_on/reviewed_on refreshed to 2026-07-29. Liveness flagged the launch command as unconfirmed by execution (auth-not-applicable but not smoke-run): fetched blazickjp/arxiv-mcp-server’s README this run — claude mcp add --transport stdio --scope user arxiv -- uvx arxiv-mcp-server and the claude_desktop_config.json stdio block both match the page verbatim. No fix needed.

Fixed

  • arrayexpress (Augmented Nature) — Pricing row corrected from “Free / OSS” to note the wrapper’s restrictive personal/non-commercial LICENSE. Liveness flagged license-missing; the GitHub license API actually returns a LICENSE file (SPDX NOASSERTION) whose text, fetched this run, is a restrictive non-commercial grant — same pattern as uniprot/alphafold/gene-ontology/human-protein-atlas. This is a license-text discrepancy the prefetch’s binary “license present/missing” signal cannot distinguish from a real absence.

Flagged

  • arrayexpress — security stays caution; added to the Flagged registry with the corrected rationale (restrictive LICENSE vs. the page’s former OSS claim, not just “no LICENSE”).

Security

  • No new advisories or provenance mismatches found in this batch. arrayexpress regrades its existing caution rationale (see Fixed/Flagged) but the grade itself is unchanged.

Note: worked the digest’s 4-page review-budget list in full; the other 9 flagged pages in this run’s 14-page liveness batch were over budget and were not touched — they stay due and lead the next worklist.

2026-07-29

Verified

  • Worked the injected 8-page worklist (allenbrain → aomic-skill) top-to-bottom. 3 pages resolved clean per the workflow’s liveness prefetch (alphafold2, amr-detection, aomic-skill) — anchors re-confirmed (google-deepmind/alphafold/science-skills Apache-2.0/2026-04-22+2026-07-07; GPTomics/bioSkills MIT/2026-07-18; CUHK-AIM-Group/NeuroClaw MIT/2026-07-14), stamps refreshed 2026-07-20→2026-07-29 with no content changes.
  • allenbrain: fetched api.github.com/repos/MCPmed/allenbrain-mcp — still resolves, no LICENSE, not archived, pushed 2026-04-01, 3 stars. Prior-run transfer fix stands; kept degraded/caution.
  • alphafold: fetched api.github.com/repos/Augmented-Nature/AlphaFold-MCP-Server — LICENSE still NOASSERTION (restrictive non-commercial), not archived, pushed 2025-12-21, 35 stars. Kept works/caution.
  • alphagenome: google-deepmind/science-skills still Apache-2.0/not-archived, pushed 2026-07-07, 2,552 stars; skill dir resolves. AlphaGenome API remains a signup-gated research preview, so kept degraded/cleared per the auth-gate rubric.
  • anndata: confirmed K-Dense-AI/scientific-agent-skills MIT/not-archived, pushed 2026-07-29 (same-day), and PyPI anndata 0.13.2 (BSD-3-Clause). The quarantined smoke job additionally reports pip install anndata as a pass this run — first smoke evidence for this page (install-only, no import/boot script) — noted in the **Verified** row. Kept works/cleared.
  • antibody-registry: refetched claude.com/docs/claude-science/connectors-and-skills — the Research Resources featured connector still lists “Grants.gov, Antibody Registry” as sources. Kept works/cleared.

Discrepancy noted (informational, no page change)

  • The injected liveness prefetch resolved antibody-registry’s “repo” to scripps-ai-enablement/sci-ai-enabler (this catalog repo) rather than an external tool repo — expected, since this is an Anthropic-hosted connector with no installable repo of its own; the page correctly documents no git-based install path. Flagging so the mismatch between prefetch and page content is visible, per this run’s instructions.

Fixed

(none this run — all 8 pages held their existing grades)

Flagged

(none new this run)

Security

No security regrades this run. All 5 adjudicated pages’ existing caution/cleared grades were reconfirmed against freshly fetched GitHub/PyPI/doc evidence (see Verified above); no new OSV/GitHub advisories found for any of the 8 repos/packages checked.

2026-07-29 (worklist maintenance — 25-page window, umap-learn → aind-data)

Worked the injected 25-page worklist top-to-bottom. All 25 were already stamped 2026-07-20; rechecked against fresh source fetches. 24 clean rechecks + 1 launch-command fix. Stamps refreshed 2026-07-20→2026-07-29.

Verified

  • 24 pages rechecked works/degraded (unchanged) against fresh anchors: K-Dense-AI/scientific-agent- skills MIT/pushed 2026-07-29/32073★ (umap-learn, usfiscaldata, vaex, venue-templates, what-if-oracle, xlsx, zarr-python, adaptyv, aeon); google-deepmind/science-skills Apache-2.0/ pushed 2026-07-07/2552★ (unibind-database — prior scienceskillscommon fix stands, stays degraded); jaechang-hits/SciAgent-Skills CC-BY-4.0-root/pushed 2026-07-24/288★ (unichem-database, uspto-database, viennarna-structure-prediction, western-blot-quantification, zinc-database); Augmented-Nature/UniProt-MCP-Server NOASSERTION/pushed 2025-12-21/19★ (uniprot); CUHK-AIM-Group/NeuroClaw MIT/pushed 2026-07-26/76★ (wmh-segmentation, abcd-skill, abide-skill, adhd200-skill, adni-skill, aibl-skill); anthropics/life-sciences marketplace 21 plugins (10x-genomics-cloud, adisinsight — both stay degraded, subscription/account-gated).
  • Added missing verification_note + **Verified** table row to unichem-database, and missing verification_note to uspto-database / viennarna-structure-prediction / western-blot-quantification.

Fixed

  • aind-data works→degraded — PyPI aind-data-mcp v0.4.5 MIT still resolves but the upstream README no longer ships a stdio console-script; it now documents a remote HTTP endpoint https://metadata-portal.allenneuraldynamics.org/mcp/ (live: 406 to a browser Accept header). No aind-data-mcp entry point in PyPI metadata. Rewrote the install block from the dead uv tool install + --transport stdio -- aind-data-mcp invocation to the current --transport http registration (plus the HTTP claude_desktop_config.json form).

Flagged

  • aind-data degraded — launch command auto-fixed to the HTTP transport this run; recheck the HTTP endpoint next cycle and flip to works once it proves stable.

Security

  • No security regrades. Kept-caution unchanged: xlsx (Anthropic-proprietary LICENSE.txt in an MIT collection), what-if-oracle (CC BY-NC-SA 4.0 non-commercial), adaptyv (writes to a paid wet-lab platform, handles ADAPTYV_API_KEY), uniprot (restrictive non-commercial LICENSE vs page’s MIT claim, ~7mo stale), 10x-genomics-cloud (closed-source vendor binary). aind-data security stays cleared (Allen Institute provenance, MIT, read-only).

2026-07-27 (worklist maintenance — 4-page recheck, latchbio-integration → liana-mcp)

Worked the injected 4-page worklist top-to-bottom. All four were already stamped 2026-07-20 (within cadence); rechecked against fresh source fetches — zero drift, no fixes. Stamps refreshed to 2026-07-27.

Verified

  • latchbio-integration works/caution (unchanged) — K-Dense-AI/scientific-agent-skills MIT/not- archived/pushed 2026-07-27/31.9k★; skills/latchbio-integration/SKILL.md (8527 B) confirmed.
  • latex-posters works/caution (unchanged) — same anchor; skills/latex-posters/SKILL.md (15554 B) confirmed.
  • lggnn works/cleared (unchanged) — CUHK-AIM-Group/NeuroClaw MIT/not-archived/pushed 2026-07-26/ 76★; skills/lggnn/SKILL.md (5467 B) confirmed.
  • liana-mcp works/caution (unchanged) — scmcphub/liana-mcp no-LICENSE/not-archived/pushed 2025-06-27/1★; PyPI liana-mcp 0.4.0; launch liana-mcp run re-confirmed against the repo README.

Security

  • latchbio-integration caution: provenance matches K-Dense-AI, but the per-skill license is unstated on the page and it deploys workflows/data to the external LatchBio cloud with user credentials.
  • latex-posters caution: provenance matches K-Dense-AI, no external calls, but the per-skill license is unstated on the page (collection root is MIT).
  • lggnn cleared: provenance matches CUHK-AIM-Group/NeuroClaw, MIT, no OSV advisories.
  • liana-mcp caution: provenance matches scmcphub, no OSV advisories, but no repo LICENSE file and the scmcphub ecosystem is unmaintained since 2025-06.

2026-07-27 (worklist maintenance — 4-page recheck, kegg-database → kmeans)

Worked the injected 4-page worklist top-to-bottom. All four were already stamped 2026-07-20 (within cadence); rechecked against fresh source fetches — zero drift, no fixes. Stamps refreshed to 2026-07-27.

Verified

  • kegg-database works/caution (unchanged) — jaechang-hits/SciAgent-Skills NOASSERTION-classifier- but-CC-BY-4.0-root/not-archived/pushed 2026-07-24/284★; skills/genomics-bioinformatics/databases/ kegg-database/SKILL.md (18418 B) confirmed.
  • kegg-pathway-analysis works/caution (unchanged) — same anchor; skills/systems-biology-multiomics/ kegg-pathway-analysis/SKILL.md (17508 B) confirmed.
  • ketcher works/cleared (unchanged) — epam/ketcher Apache-2.0/not-archived/pushed 2026-07-27/847★; npm ketcher-react 3.17.1 Apache-2.0; Anthropic Claude Science connector, client-side editor, no creds.
  • kmeans works/cleared (unchanged) — CUHK-AIM-Group/NeuroClaw MIT/not-archived/pushed 2026-07-26/ 76★; skills/kmeans/SKILL.md (4082 B) confirmed.

Security

  • kegg-database + kegg-pathway-analysis caution: skill code CC BY 4.0 and provenance clear, but the underlying KEGG data needs a paid commercial license for non-academic use (data-use restriction only).
  • ketcher + kmeans cleared: provenance matches supplier, real Apache-2.0/MIT licenses, no OSV advisories, no credentials.

2026-07-27 (worklist maintenance — 4-page recheck, inductive-bio → intact)

Worked the injected 4-page worklist top-to-bottom. All four were already stamped 2026-07-20 (within cadence); rechecked against fresh source fetches — zero drift, no fixes. Stamps refreshed to 2026-07-27.

Verified

  • inductive-bio degraded/cleared (unchanged) — supplier inductive.bio loads (Beacon-1 ADMET, 3× OpenADMET blind-challenge wins) but no public MCP endpoint / self-serve sign-up, and the Claude Science connectors-and-skills doc does not list Inductive Bio; the vendor connector is confirmed only via the cited PR Newswire release, so the install path stays functionally unverifiable.
  • infographics works/caution (unchanged) — K-Dense-AI/scientific-agent-skills MIT/not-archived/ pushed 2026-07-27/31.9k★; skills/infographics/SKILL.md (10949 B) + references/scripts confirmed.
  • instrument-data-to-allotrope works/cleared (unchanged) — plugin registered in anthropics/ life-sciences .claude-plugin/marketplace.json (source ./); allotropy dep on PyPI is Benchling MIT (latest 0.1.142).
  • intact works/cleared (unchanged) — Claude Science connectors-and-skills doc confirms “Structures & Interactions | PDB, AlphaFold, EMDB, Complex Portal, IntAct”; read-only EMBL-EBI CC BY 4.0.

Security

  • infographics caution: unstated skill license + prompts sent to external Google Gemini / Nano Banana image services. The other three unchanged (inductive-bio cleared vendor connector; allotrope/intact cleared Anthropic-provenance, real MIT/CC-BY licenses, no OSV advisories).

2026-07-27 (worklist maintenance — 4-page recheck, hypothesis-crucible → ica)

Worked the injected 4-page worklist top-to-bottom. All four were already works/cleared (stamped 2026-07-20, within cadence); rechecked against fresh source fetches — zero drift, no fixes. Stamps refreshed to 2026-07-27. Catalog is now at full verification: coverage (459 total · 0 unstamped).

Verified

  • hypothesis-crucible works/cleared — first-party this repo; confirmed crucible/.claude-plugin/ plugin.json (name crucible), crucible/skills/forge/SKILL.md, and the crucible entry in root .claude-plugin/marketplace.json so /plugin install crucible matches.
  • hypothesis-generation works/cleared — K-Dense-AI/scientific-agent-skills MIT/not-archived/ pushed 2026-07-27/31.9k★; skills/hypothesis-generation/ holds SKILL.md (14767 B) + assets/ references/scripts.
  • ibgnn works/cleared — CUHK-AIM-Group/NeuroClaw MIT/not-archived/pushed 2026-07-26/76★; skills/ibgnn/SKILL.md (5587 B) resolves.
  • ica works/cleared — same NeuroClaw anchor; skills/ica/SKILL.md (4275 B) resolves.

Security

  • All four cleared: provenance matches supplier, real licenses (MIT / first-party OSS), no OSV advisories, maintained. Skills execute locally via Bash/Python (no smoke target).