Install
$ agentstack add skill-amey-thakur-ai-skills-vendor-evaluation ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Vendor evaluation
Choosing a vendor is choosing a dependency you will live with for years and pay to leave. The evaluation goes wrong when it is run on demos and vibes: the polished sales walkthrough decides it, criteria get invented to justify the favorite, and no one asks what it costs to get back out. A disciplined evaluation tests the vendor on your work and prices the exit before it prices the entry.
Method
- Set weighted criteria before you see a single demo. List what matters:
functionality, reliability, security posture, support, total cost, and how hard it is to leave. Assign weights that sum to a fixed total, and write them down first. Criteria invented after the demo just rationalize the vendor who demoed best.
- Score against requirements, not against each other. Rate every vendor on
each criterion against your bar, so a weak field cannot make a mediocre option look strong by comparison. Separate must-haves, which are pass or fail, from nice-to-haves that earn weighted points.
- Design the proof of concept around your real workload. Feed it your actual
data shapes, your peak volume, your awkward edge cases, and your integration points. A POC on the vendor's sample data proves their demo works. Define the success metrics and the time box before you start so the POC ends in a verdict, not an open-ended trial.
- Cost the exit, not just the entry. Estimate what it takes to migrate off:
data export formats, proprietary interfaces you would rewrite, retraining, and contract lock-in. High switching cost is a real price even if it never appears on an invoice. Prefer vendors whose data you can get back in a usable form.
- Verify security and compliance with evidence, not the questionnaire alone.
Ask for the SOC 2 or ISO report, the data processing terms, subprocessor list, breach history, and where data is stored. A vendor who cannot produce these for a serious deal is telling you something.
- Reference-check past the list they gave you. Talk to the references, then
find a customer they did not hand you and ask what breaks at month six: support response times, hidden costs, the feature that was "on the roadmap." Sales sells the roadmap; a real customer tells you the product.
Checks
- Were the weighted criteria fixed before demos, or shaped to fit a favorite?
- Did the POC run on your workload and data, or on the vendor's canned scenario?
- Can you state, in dollars and weeks, what leaving this vendor would cost?
Boundaries
This picks the vendor; it does not negotiate the contract terms or run procurement and legal review, which have their own owners and thresholds. Approval limits and required security sign-offs are company policy: route spend and data agreements through the process your organization requires rather than deciding them here.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Amey-Thakur
- Source: Amey-Thakur/AI-SKILLS
- License: MIT
- Homepage: https://amey-thakur.github.io/AI-SKILLS/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.