CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

tooluniverse-precision-oncology

Cancer treatment recommendations from molecular profile (mutations + cancer type + biomarkers) — FDA-approved + investigational therapies, resistance mechanisms, matching clinical trials, prognosis. Uses CIViC, ClinVar, OpenTargets, ClinicalTrials.gov. Use for tumor-board treatment recommendations, evidence-tiered actionability assessment, and FDA-precedent-driven therapy selection.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

The canonical home for this skill is tooluniverse-precision-oncology in mims-harvard/ToolUniverse

SKILL.md
Quality
Evals
Security

Quality

Content

67%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 a lean, highly actionable operator's guide with good phase sequencing and effective parameter-gotcha coverage, but its progressive disclosure is fundamentally broken: all six referenced detail files are missing from the bundle, and the one bundled script is orphaned. The strongest fix is to ship the referenced files (or remove the dead links) and surface the GDSC script.

Suggestions

Ship the six referenced files (TOOLS_REFERENCE.md, API_USAGE_PATTERNS.md, TREATMENT_ALGORITHMS.md, REPORT_TEMPLATE.md, EXAMPLES.md, CHECKLIST.md) or delete the References section — dead links break every path from the overview to the detail layer.

Move the biomarker-to-drug mapping blocks and resistance-mutation frequency tables out of SKILL.md into TREATMENT_ALGORITHMS.md (which already claims to cover 'cancer type mappings'), keeping only a pointer in the body.

Reference scripts/gdsc_drug_response.py from the body — it is the only real bundle file, covers GDSC drug-sensitivity data the workflow otherwise lacks, and is currently undiscoverable.

DimensionReasoningScore

Conciseness

The body is dense and operational throughout — tool names, exact parameters ('ensemblId, camelCase', 'condition, NOT disease', 'variant_id (numeric, e.g., 4170)'), a WRONG/CORRECT table, and domain facts Claude cannot reliably recall (resistance mutation frequencies, OncoKB demo-mode limits). It stays at 4 rather than 5 because the ~10-line biomarker-to-drug mapping blocks duplicate content the References section says TREATMENT_ALGORITHMS.md already contains ('cancer type mappings'), which is inlining that could be trimmed.

4 / 5

Actionability

Guidance is highly concrete: exact MCP tool names per phase, parameter gotchas with example values ('molecular_profile="EGFR C797S"'), and 'LOOK UP DON'T GUESS' rules naming the exact tool to call per fact type. It falls short of the copy-paste-ready 5-anchor because the body shows no literal example tool calls or code — those are deferred to API_USAGE_PATTERNS.md — and several phases list tools without stating the query to issue.

4 / 5

Workflow Clarity

A clear phase sequence (Phase 0 verification → Phase 6 report) with per-phase tool lists, a tool-verification table, mandatory checkpoints ('You MUST call FAERS... before finalizing the report'), and fallback guidance (OncoKB demo mode → use CIViC). It does not reach 5 because per-phase validation/feedback loops are implicit rather than explicit — the checklist lives in CHECKLIST.md rather than as inline verify steps, and there is no stated error-recovery path when a lookup returns nothing.

4 / 5

Progressive Disclosure

The References section signals six one-level-deep files (TOOLS_REFERENCE.md, API_USAGE_PATTERNS.md, TREATMENT_ALGORITHMS.md, REPORT_TEMPLATE.md, EXAMPLES.md, CHECKLIST.md), but none of them exist in the bundle — every link is dead, so the entire detail layer is unreachable. Additionally, content that the references say belongs in those files (treatment mappings, examples) is inlined, and the one real bundle file, scripts/gdsc_drug_response.py, is never referenced from the body. This matches the 2-anchor (inlined content that belongs in separate files; navigation broken) rather than the 3-anchor, which assumes references are at least present.

2 / 5

Total

14

/

20

Passed

Description

92%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.

A strong description that answers both what the skill does and when to use it, with concrete capability enumeration and explicit trigger guidance. Its only weakness is modest synonym coverage for natural user phrasings of the triggers.

DimensionReasoningScore

Specificity

The description enumerates multiple concrete capabilities: 'Cancer treatment recommendations from molecular profile (mutations + cancer type + biomarkers)', 'FDA-approved + investigational therapies, resistance mechanisms, matching clinical trials, prognosis', and names the data sources used (CIViC, ClinVar, OpenTargets, ClinicalTrials.gov). This matches the top anchor (multiple specific concrete actions, comprehensive coverage); it is well above the 4-anchor, which requires 'minor gaps in coverage' — the what-to-output surface (therapies, trials, resistance, prognosis) is fully enumerated.

5 / 5

Completeness

Both halves are explicit: the 'what' is the full capability list up front, and the 'when' is an explicit trigger clause — 'Use for tumor-board treatment recommendations, evidence-tiered actionability assessment, and FDA-precedent-driven therapy selection.' This mirrors the 5-anchor example structure ('Extract... Use when...'). It is clearly above the 4-anchor, where the 'when' is present but could be more specific — here three distinct use cases are named.

5 / 5

Trigger Term Quality

Natural terms users would say are present: 'treatment recommendations', 'mutations', 'cancer type', 'biomarkers', 'clinical trials', 'tumor-board'. However, common synonyms a patient-facing query might use are missing — 'targeted therapy', 'drug resistance', 'precision medicine', 'immunotherapy', 'second-line therapy' — so it fits the 4-anchor ('good keyword coverage; a few natural terms missing') rather than the comprehensive 5-anchor.

4 / 5

Distinctiveness Conflict Risk

It carves a clear niche (cancer therapy selection from molecular profile) with distinct trigger phrases ('tumor-board', 'evidence-tiered actionability', 'FDA-precedent-driven therapy selection') and named genomics databases, so it is unlikely to fire for unrelated biomedical skills. It sits above the 4-anchor because the overlap risk with closely related skills (e.g., a drug-interaction skill) is minimal given the cancer-treatment framing throughout.

5 / 5

Total

19

/

20

Passed

Validation

93%

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

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

relative_links

Relative link issues: 6 missing

Warning

Total

15

/

16

Passed

Repository

mims-harvard/ToolUniverse

Path

plugins/tooluniverse/skills/tooluniverse-precision-oncology/SKILL.md

Reviewed

Table of Contents

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.