- Home
- Skills
- Agents & Orchestration
- ποΈ Model Inventory Auditor
More screenshots
Works with the AI Agents you already use
ποΈ Model Inventory Auditor
Find pattern-based model/provider references in reviewed source and prepare an explicitly incomplete inventory for deployment review.
$13
ποΈ Model Inventory Auditor
Example session with this skill installed
For a separately obtained complete download, run the packaged risky-sample fixture with Model Inventory Auditor v1.1.2. Explain the findings without treating detected strings as complete runtime inventory.
- Read your context and instructions
- Compiled the model inventory auditor
Local v1.1.2 fixture run, not Agensi delivery or customer evidence
Command: python scripts/scan_model_inventory.py fixtures/risky-sample
Coverage
incomplete. The run detects OpenAI-associated model strings gpt-4o at app.py:6 and gpt-4o-2024-08-06 at app.py:7. Paths below are package-relative.
All seven reported findings
- MIA001, medium, app.py:6: model-like literal gpt-4o to review.
- MIA004, medium, app.py:6: review configuration indirection.
- MIA001, medium, app.py:7: model-like literal gpt-4o-2024-08-06 to review.
- MIA004, medium, app.py:7: review configuration indirection.
- MIA002, high, summary: only one model provider detected, OpenAI. This does not prove no fallback exists.
- MIA003, medium, summary: logical model gpt-4o has inconsistent IDs.
- MIA005, low, package.json:1: manifest declares provider dependency; runtime use and alternatives require review: OpenAI.
Next step
Review deployment identity and runtime use. The optional JSON handoff starts incomplete with unresolved provider/platform identity; do not convert these detections into a complete bill of materials or a retirement verdict.
Connects securely to your tools. The creator never sees your data.
About this skill
What you get
Model Inventory Auditor v1.1.3 provides a read-only scanner, editable detection patterns, fixtures, tests and an optional incomplete JSON retirement handoff. Findings include model strings and file/line references, not copied source lines.
Who it helps
A developer or maintainer starting a model-dependency review. Version 1.1.3 separates supported model-selection patterns from ambiguous prose, comments, YAML documentation, negative declarations and JSON pricing records. Generic identifier patterns remain review-only. A single detected provider does not prove that the application lacks a fallback.
Before you start
Agensi provides agent-connected access, not a hosted app or ZIP download. Python 3.10+ is required; use Python 3.11+ for pyproject.toml. The connected host must retrieve the complete scanner and provider-rules reference and execute Python. The full entrypoint accepts complete rules in memory without saving skill assets or patching the scanner. Direct candidate-host testing does not establish fresh Agensi delivery or compatibility with every host. For a separate download, extract the complete folder. Supply reviewed non-secret inputs. Environment files, including .env.example, are excluded; other secrets are not reliably detected.
First run
Ask your agent to retrieve the complete scanner and rules, then follow the Agent-connected execution section in SKILL.md. Start with synthetic fixtures. For the separate complete download, run python scripts/scan_model_inventory.py fixtures/risky-sample. Add --json for the incomplete retirement handoff.
Limits
No complete runtime discovery, model calls, package installation or source edits. Dynamic routing, unsupported providers, ambiguous selection styles and skipped content remain coverage gaps. Conservative filtering can undercount legitimate model selections; pricing structures outside JSON need manual review. Missing or invalid rules, missing inputs and unreadable inputs return CANNOT_ASSESS, not a clean assessment.
Pattern-based and incomplete. Does not establish runtime dependency, complete coverage, absence of fallback, or retirement applicability. Environment files are excluded; review all other input for secrets. Dynamic routing, unsupported providers and skipped content need separate review. Missing, linked or unreadable input cannot produce a clean assessment.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 25 days ago
Requires
Permissions
File Scopes
Read-only scanner. Do not execute scanned project code. Environment files are excluded; provide non-secret inputs only. Retrieve the complete provider-rules reference. The documented entrypoint accepts complete rules in memory. No network or environment-variable access.