molecular-rag

Retrieve structurally similar compounds with known properties from ChEMBL/ZINC to ground predictions and inform optimization. Based on MolRAG (Xian 2025, ACL).

Quality

65%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/chemistry/molecular-rag/SKILL.md

Quality

Content

68%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is well-structured and actionable with copy-paste-ready commands backed by a real script, and progressive disclosure is handled cleanly with a one-level-deep script reference. The main gap is the absence of validation/verification checkpoints for a batch API-retrieval workflow.

Suggestions

Add a verification step to each workflow, e.g. after running retrieve_analogs.py check that analogs.json has num_results > 0 and that top similarities meet the threshold before using them to ground a prediction.

Add a short 'Interpreting results' example showing how to read analogs.json (similarity, bioactivity values) and feed it back into a property prediction, closing the RAG loop.

Trim the Overview's motivational hERG example to one line or move it under When to Use to improve token efficiency.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient with concrete bash commands and a tight script-reference table, but the Overview includes light motivational padding ("LLMs hallucinate molecular properties" and the illustrative hERG sentence) that could be trimmed without losing clarity.	4 / 5
Actionability	Three copy-paste-ready bash workflows with full flags, SMILES examples, and output paths cover the common cases, and the referenced script exists, but there is no example showing how to consume analogs.json to actually ground a prediction.	4 / 5
Workflow Clarity	Each workflow is a clear single command, but this is a batch operation (retrieving and enriching up to 20 compounds via a rate-limited API loop) with no validation/verification checkpoint in SKILL.md, which caps workflow clarity at 3 per the rubric.	3 / 5
Progressive Disclosure	SKILL.md is a well-organized overview (Overview, When to Use, Installation, Core Workflows, Script Reference) pointing one level deep to the real scripts/retrieve_analogs.py, but the output JSON schema and ChEMBL API patterns are not split into a reference file, leaving a minor organization gap.	4 / 5
	Total	15 / 20 Passed

Description

61%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description clearly communicates what the skill does and its domain, with good trigger-term coverage, but lacks an explicit 'Use when' clause and could be more comprehensive in listing concrete actions. It is mostly distinct from sibling skills with minor overlap risk.

Suggestions

Add an explicit 'Use when...' clause naming natural triggers, e.g. 'Use when predicting molecular properties, optimizing a lead scaffold, or assessing a generated molecule's novelty via analog search.'

Include common synonyms users say ('analogs', 'analogs search', 'similarity search') alongside 'structurally similar compounds' to improve trigger-term coverage.

List 1-2 more concrete actions (e.g. 'rank analogs by Tanimoto similarity', 'fetch experimental bioactivities') to lift specificity toward comprehensive coverage.

Dimension	Reasoning	Score
Specificity	The description names the domain (compounds from ChEMBL/ZINC) and one concrete action ("Retrieve structurally similar compounds with known properties"), but "ground predictions and inform optimization" are purposes rather than additional concrete actions, so coverage is not comprehensive.	3 / 5
Completeness	The 'what' is clearly stated (retrieve similar compounds with known properties), but there is no 'Use when...' clause or equivalent explicit trigger guidance, so completeness is capped at 3 per the rubric guideline.	3 / 5
Trigger Term Quality	It includes natural domain terms a medicinal chemist would say ("structurally similar compounds", "ChEMBL", "ZINC", "ground predictions"), but omits common synonyms users actually say like "analogs", "analogs search", or "similarity search".	4 / 5
Distinctiveness Conflict Risk	The niche is fairly specific (molecular RAG over ChEMBL/ZINC for grounding predictions), but there is minor overlap risk with closely related database skills (chembl-database, zinc-database) that also retrieve compounds from these sources.	4 / 5
	Total	14 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/chemistry/molecular-rag/SKILL.md
Commit: 3a6c3a9

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.