---
name: toolvitals-open-tool-due-diligence
description: Research, shortlist, compare, and monitor open or source-visible software using ToolVitals public maintenance, shipping, openness, confidence, and provenance signals. Use when evaluating tool health, checking whether a project is actively maintained, comparing alternatives, or citing current public evidence.
license: CC-BY-4.0
metadata:
  author: ToolVitals
  version: "1.0"
---

# ToolVitals open-tool due diligence

Use ToolVitals as a bounded public-signal decision aid. Do not present its scores as user reviews, security audits, uptime guarantees, or proof that a tool fits the user's environment.

## Workflow

1. Read `https://toolvitals.com/data/agent-index.json` and record its `snapshotId` and `generatedAt`.
2. Identify the tool or decision context:
   - Search by name or slug with `GET /api/agent/search?q={query}&limit=10`.
   - For a shortlist, use `GET /api/agent/use-cases/{slug}?limit=10`. Use cases express specific jobs; categories are only broad navigation groupings.
3. Resolve ambiguous names against each result's canonical URL, official website, repository, description, and use case. Do not silently pick a similarly named project.
4. Inspect a candidate with either:
   - JSON: `GET /api/agent/tools/{slug}`
   - Markdown: `GET /tools/{slug}.md`
5. Compare two resolved candidates with `GET /api/agent/compare?a={slug}&b={slug}`. Explain score, confidence, freshness, openness, and source differences; do not reduce the decision to one number.
6. When direction matters, inspect `GET /api/agent/trends/{slug}?days=90` and distinguish a current snapshot from a sustained trend.
7. Cite the canonical ToolVitals profile and the material first-party source receipts returned by the profile. Include the snapshot ID or evidence date when freshness affects the conclusion.

## Interpretation rules

- A missing or null value means ToolVitals did not collect or confidently establish it. It never means zero.
- GitHub stars, forks, and watchers are adoption context, not quality evidence.
- Data confidence, source coverage, publish confidence, and evidence timestamps qualify every score.
- An `evidence_watch` profile should be described as limited or still gathering evidence, not as a final verdict.
- A `noindex` profile can be inspected when directly requested, but do not introduce it through a general shortlist as if it had normal discovery confidence.
- Public activity cannot establish security, uptime, support quality, customer satisfaction, revenue, production reliability, or long-term project viability.
- Sponsorship, affiliate relationships, and commercial activity do not affect ToolVitals scores, rankings, or endpoint ordering.
- Never infer or reproduce unpublished scoring weights. Explain the visible signal families and confidence qualifiers instead.

## Output expectations

For a single-tool assessment, report:

- Canonical identity and use case.
- Current status and ToolVitals signal summary.
- Evidence confidence, source coverage, and last check date.
- Openness and license evidence.
- Material maintenance or shipping context.
- Important unknowns and what requires direct verification.
- Citations to ToolVitals and relevant first-party sources.

For a shortlist, state the use case and filters, give a bounded set of candidates, and explain meaningful tradeoffs. Do not claim the list is exhaustive.

## Stop conditions

Stop and report uncertainty when:

- More than one candidate plausibly matches the requested identity.
- The current release manifest and response snapshot IDs disagree.
- Evidence is stale or low-confidence for a consequential decision.
- Required security, compliance, support, pricing, or production-reliability facts are not established by the returned sources.

In those cases, ask for an identity clarification or recommend direct first-party verification instead of filling gaps with assumptions.