tooluniverse-chemical-compound-retrieval

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha 43be0fac747b · 31 Oct 2026Report an issue or request a re-scan

• Publishermims-harvard GitHub organizationOwner of the repository this skill is read from
• Source foldermims-harvard/tooluniverse/skills/tooluniverse-chemical-compound-retrievalThe exact folder we audited
• LicenceApache-2.0As the repository states it
• Commit pinned for this verdict43be0fac747bA new audit runs when the source changes
• Repository statusNot archivedLast push 3 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Retrieve chemical compound data from PubChem and ChEMBL with disambiguation, cross-referencing, and stereochemistry handling. Use for resolving compound names to SMILES/InChI/CID/ChEMBL IDs (including OPSIN…

---
name: tooluniverse-chemical-compound-retrieval
description: Retrieve chemical compound data from PubChem and ChEMBL with disambiguation, cross-referencing, and stereochemistry handling. Use for resolving compound names to SMILES/InChI/CID/ChEMBL IDs (including OPSIN deterministic IUPAC-name-to-structure parsing), fetching molecular properties, distinguishing isomers/stereo forms, and cross-validating identity across databases. Always use English compound names; flags ambiguous queries (e.g., Vitamin D has multiple forms).
disable-model-invocation: true
---

# Chemical Compound Information Retrieval

Retrieve comprehensive chemical compound data with proper disambiguation and cross-database validation.

**LOOK UP DON'T GUESS**: Never assume a CID, ChEMBL ID, or molecular property value. Always retrieve from PubChem/ChEMBL.

**English-first**: Always use English compound names in tool calls. Respond in user's language.

## Domain Reasoning: Disambiguation

"Aspirin" = one compound. "Vitamin D" = multiple forms (D2/D3/active metabolite). For generic class names (steroids, vitamins, acids), present candidates and confirm before proceeding.

---

## Workflow

```
Phase 0: Clarify (only if highly ambiguous -- skip for unambiguous names or specific IDs)
Phase 1: Disambiguate → resolve PubChem CID + ChEMBL ID
Phase 2: Retrieve data (silent)
Phase 3: Report compound profile
```

### Phase 1: Disambiguation

```python
# By name
result = tu.tools.PubChem_get_CID_by_compound_name(compound_name=name)
# By SYSTEMATIC (IUPAC) name -> structure, deterministic parser (no DB lookup)
opsin = tu.tools.OPSIN_name_to_structure(name="2-acetoxybenzoic acid")
# Returns {parsed, smiles, inchi, inchikey}; use the SMILES/InChIKey to anchor a
# PubChem_get_CID_by_SMILES lookup. Trade/trivial names give parsed=false -> fall
# back to PubChem_get_CID_by_compound_name for those.
# By SMILES
result = tu.tools.PubChem_get_CID_by_SMILES(smiles=smiles)
# Cross-reference
chembl_result = tu.tools.ChEMBL_search_molecules(query=name, limit=5)
```

Verify: CID + ChEMBL ID + canonical SMILES + stereochemistry + salt forms.

### Phase 2: Data Retrieval

**PubChem**: `PubChem_get_compound_properties_by_CID`, `PubChemBioAssay_get_assay_summary`, `PubChemTox_get_acute_effects`, `PubChem_get_compound_2D_image_by_CID`

**ChEMBL**: `ChEMBL_get_compound_record_activities`, `ChEMBL_get_molecule_targets`, `ChEMBL_get_assay_activities`

**Optional**: `PubChem_get_associated_patents_by_CID`, `PubChem_search_compounds_by_similarity`

### Phase 3: Report

Compound Profile with: Identity (CID, ChEMBL ID, IUPAC, SMILES), Chemical Properties (MW, LogP, HBD, HBA, PSA, Lipinski), Bioactivity (targets, IC50/Ki), Drug Info (if approved), Data Sources.

---

## Fallback Chains

| Primary | Fallback |
|---------|----------|
| PubChem name lookup (systematic name) | `OPSIN_name_to_structure` → SMILES/InChIKey → PubChem_get_CID_by_SMILES |
| PubChem name lookup | ChEMBL search → SMILES → PubChem_get_CID_by_SMILES |
| ChEMBL bioactivity | PubChem bioassay summary |
| Drug label | Note "unavailable" |

---

## Evidence Grading

| Grade | Criteria |
|-------|----------|
| **Confirmed** | CID + ChEMBL cross-match, InChI/SMILES agree |
| **Probable** | CID found, partial ChEMBL match |
| **Uncertain** | Single database only, or multiple CIDs |
| **Unverified** | No cross-reference, single-source |

**Bioactivity**: ChEMBL > PubChem BioAssay for curated data. IC50/Ki < 100nM = potent, 100nM-1uM = moderate, >10uM = weak. Lipinski violations reduce oral bioavailability but don't disqualify.

---

## SMILES Verification

Always verify novel SMILES: `python3 src/tooluniverse/tools/smiles_verifier.py --smiles "SMILES_STRING"`. Invalid SMILES produce wrong results or cryptic errors.

---

## Identifier Resolution and Cross-Database Referencing

Beyond PubChem/ChEMBL, two more tools resolve identifiers when a name doesn't
hit in PubChem or you need cross-references across other chemical databases:

- **`NCICACTUS_resolve`** (NCI CACTUS resolver) — converts between chemical
  names (IUPAC or common), SMILES, InChI, InChIKey, CAS number, and molecular
  formula. Verified live: `NCICACTUS_resolve(identifier="ibuprofen", representation="smiles")`
  → `CC(C)Cc1ccc(cc1)C(C)C(O)=O`; also resolves a CAS number (`"50-78-2"`) straight
  to an IUPAC name (`"2-acetyloxybenzoic acid"`, i.e. aspirin) or an InChIKey to
  a formula. Useful as a second name-resolution path alongside `OPSIN_name_to_structure`
  and `PubChem_get_CID_by_compound_name` — CACTUS handles CAS numbers and trivial
  names OPSIN can't parse.
- **`UniChem_search_compound`** / **`UniChem_connectivity_search`** — given an
  InChIKey (or a source-database compound ID), returns cross-references across
  UniChem's 25+ registered source databases (ChEMBL, DrugBank, PubChem, Wikipedia,
  ChEBI, etc. — verified live via `UniChem_list_sources`). `connectivity_search`
  additionally finds compounds sharing the same connectivity layer (same skeleton,
  different stereochemistry/salt/isotope) — use this when you want "the same
  molecule under any name/form" rather than an exact match. Verified live on
  aspirin's InChIKey (`BSYNRYMUTXBXSQ-UHFFFAOYSA-N`): both returned real
  cross-database records including a Wikipedia link.

**Chained identifier workflow (real values):** `NCICACTUS_resolve(identifier="aspirin", representation="smiles")`
→ `CC(=O)Oc1ccccc1C(O)=O` → resolve the same identifier with `representation="stdinchikey"`
→ real result `"InChIKey=BSYNRYMUTXBXSQ-UHFFFAOYSA-N"` (**note the `InChIKey=` prefix** —
verified live; strip it before passing the bare key to another tool) → feed the
bare `BSYNRYMUTXBXSQ-UHFFFAOYSA-N` into `UniChem_search_compound` for cross-database
IDs → `PubChem_get_CID_by_SMILES` to re-anchor in PubChem if needed.

## Cheminformatics Calculations and Structure Visualization

For calculations and rendering beyond property lookup:

- **`RDKit_pharmacophore_features`** — extracts pharmaco

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/mims-harvard/tooluniverse
git -C tooluniverse checkout 43be0fac747b

Works in

1. 2026-10-31
   ✓SAFE B 89 / 100

   first audit · source 43be0fac747b

Questions about tooluniverse-chemical-compound-retrieval

• github.com/mims-harvard/tooluniverse/tree/main/skills/tooluniver — the audited source at commit 43be0fac747b and every finding aboveObserved · 31 Oct 2026

Spotted something wrong? Report a correction · Methodology