Best fit
A portfolio owner needs a release-impact retest queue or a retirement assessment against reviewed deployment records.
// AI Agents & LLM Ops
Model Release Impact Mapper
Turn sourced model-release or retirement notices into a scoped review queue: affected workflows, excluded deployment routes, unknowns and deadlines.
Access: Agent-loaded on Agensi, not a standalone app. Downloadable access applies only where a separate download offer is linked. Package version 1.1.0.
Package requirements: Python 3.10 or newer; standard library only. A terminal, or an agent that can read SKILL.md and run local Python. For a downloadable package, extract the complete folder before use.
You get: Existing release-impact mapper plus a scoped retirement mapper, input contracts, fixtures and tests.
Licence and purchase details
Review the package licence and seller terms at checkout. An AI-agent subscription or setup service is not included.
Agensi access: Execution of bundled scripts through its new connector is not yet verified. Local setup examples apply to separately obtained downloadable packages.
New in 1.1.0: scoped retirement review
Already have a reviewed deployment inventory? Use the free mapper directly. Pair exact platform, provider and model identities with dated retirement notices using references/retirement-contract.md.
python scripts/map_retirements.py INPUT.json --output-dir NEW_DIRECTORYReplace INPUT.json with your contract-shaped input and choose a new output directory. The report distinguishes APPLIES, DOES_NOT_APPLY and UNKNOWN_SCOPE, separates announcement dates from effective deadlines, and preserves incomplete-inventory warnings.
A deadline HOLD does not mean a regression occurred. Suggested alternatives remain untested and the regression handoff stays not_run. The existing release-impact mode and its walkthrough below are unchanged.
What problem does Model Release Impact Mapper solve?
A new model, agent platform, or protocol release can invalidate compatibility claims, pricing tables, output contracts, permission assumptions, and test fixtures across a portfolio before the owner knows which products are exposed.
Use it to
- Triage a skill catalog after a major US or non-US model release
- Identify MCP clients, servers, gateways, and claims exposed to a protocol revision
- Find cost tables and pricing claims affected by dated provider-rate changes
- Build a bounded retest queue before editing compatibility copy or launching provider-specific work
What it returns
- Deterministic CLEAR, REVIEW, or HOLD portfolio posture
- Prioritized item-by-change retest queue with stable scores and finding codes
- Fact-versus-inference findings for affected claims, artifacts, permissions, and runner evidence
- Revalidation coverage metrics and a deterministic SHA-256 receipt
A representative input and result
input DeepSeek V4-Pro adds Responses API support, flexible reasoning effort, and dated peak/off-peak pricing. The manifest contains a live cost reconciler plus a staged agent gateway with declared API, pricing, and reasoning exposures.
result Posture: HOLD. Three item-change pairs require revalidation: the cost reconciler pricing window is URGENT, the gateway reasoning-effort contract is HIGH, and the cost reconciler Responses API fixture is HIGH. Findings separate documented release facts from inferred impact and claim no regression.
// 30-second fixture-backed walkthrough
What does a release-impact map look like?
The included DeepSeek V4-Pro fixture moves from three documented release changes to a prioritized revalidation queue. HOLD means evidence is still required; it does not prove that a regression occurred.
Open the 30-second demo on YouTube.
Read the 30-second transcript
A new model shipped. So, what should we retest?
In the DeepSeek V4-Pro example, three things changed: Responses API support, pricing, and reasoning-effort modes.
One local command maps release facts to exposure. No provider call or credentials.
The result is HOLD: one urgent pricing check and two high-priority revalidations. None are resolved yet.
This queue does not prove regression. The free mapper is available on JustHandled Labs and Agensi for testing.
Access and approval boundaries
- Terminal permission to run the bundled local Python mapper.
- Read access to one user-supplied UTF-8 JSON release and portfolio packet.
- Write access only to the selected local JSON and Markdown report paths.
- No browser, network, API key, credential, account, model-call, marketplace, or publication permission is required.
Known limitations
- The mapper relies on owner-supplied release facts and declared portfolio domains; it does not browse or authenticate the underlying sources.
- A mapped impact is a retest obligation, not evidence that a regression occurred.
- CLEAR covers only the supplied manifest and evidence. Omitted dependencies, claims, and products remain unassessed.
- Semantic tool, authorization, agent-control, and protocol checks require both a compatible or sealed runner status and runner-contract evidence.
- The package does not measure market demand, run paid model calls, edit products, change prices, or publish updates.
// choose with context
Is Model Release Impact Mapper the right skill?
It returns
A release-impact queue, or a scoped retirement report with applicability, deadlines, unknowns and a regression handoff marked not_run.
Do not use it as
Not source authentication, runtime discovery, model testing, an automatic migration, or proof of cross-model compatibility.
Compatibility, access, version, and licence
- Reads or accesses
- User-supplied local UTF-8 JSON release/portfolio records, or retirement notices and deployment records following the included contracts. Writes only the selected local reports.
- Compatibility
- Python 3.10 or newer; standard library only. A terminal, or an agent that can read SKILL.md and run local Python. For a downloadable package, extract the complete folder before use.
- Version
- 1.1.0
- Licence
- Review the package licence and seller terms at the linked destination.
Questions
Does this run tests against the new model?
No. It maps documented release changes to declared portfolio exposure and tells you what needs testing. Actual model calls and semantic evaluation remain separate.
How is this different from an agent regression suite?
A regression suite scores a chosen agent or skill against cases. This mapper works one step earlier by identifying which products, claims, fixtures, and permission boundaries should enter that suite after a release.
Can it cover non-US model providers?
Yes. Provider and region are factual fields, not risk scores. The included fixture uses DeepSeek and keeps geography separate from severity and evidence.
Does HOLD prove one of my products is broken?
No. HOLD means a live high-criticality product has an unresolved high-severity revalidation obligation or a recorded check failed. It does not invent a regression result.
What happens when a runner contract is missing?
The report preserves CANNOT_ASSESS for semantic tool, authorization, agent-control, and protocol work. It does not build an adapter or claim the check passed.
Does it change compatibility copy or prices?
No. It returns a prioritized evidence queue only. Every product, listing, price, and publication decision remains with the owner.
Can I compare repeated runs?
Yes. Normalized inputs and report cores receive deterministic SHA-256 receipts, and stable finding codes make release-to-release comparison practical.
Model Release Impact Mapper keeps proof and approval boundaries visible.
The listing includes the tested package, realistic samples, declared permissions, and known limitations.
Get Model Release Impact Mapper on Agensi