AI inventory
List declared uses, data categories, evidence and open fields. For agent use, the file also records who authorized it, what it may do, until when, and who can shut it off.
No member data, vendor ranking, certification or savings claim.
Evidence Desk
Six tools prepare the six-column evidence file from declared inputs. Your reviewer keeps the decision.
Six declared-input preparation tools.
Not listed. No public install URL.
Examiner question preparation
Map supplied answers and declared references to eight CSBS questions. A supplied answer remains separate from inspected or sufficient evidence.
No member data, vendor ranking, certification or savings claim.
Renewal terms
Compare confirmed quote mappings on the same activity. Keep totals beside unknown fees and terms.
No member data, vendor ranking, certification or savings claim.
Procedure version
Retain supplied versions, the declared controlling-source basis and the question still open.
No member data, vendor ranking, certification or savings claim.
Pilot acceptance
Keep required cases, failed attempts and missing cases. A later pass on a different instruction leaves the original case unresolved.
No member data, vendor ranking, certification or savings claim.
Core-change reach
Independent of Jack Henry. Compare supplied files for heuristic candidate dependents and retest questions. No compilation, execution or core connection.
No member data, vendor ranking, certification or savings claim.
Prepare a review in Claude Code.
Six skills use the same guarded methods: AI inventory, pilot acceptance, agent authority, renewal terms, procedure version and core-change reach.
Download the local Claude evaluation package (5.6 MB). Node 24 or later is required. The package includes the local server and its locked dependencies.
- Extract the archive and keep the marketplace folder intact.
- Register that local folder and install Evidence Desk in an institution-approved Claude Code environment.
- Choose a review skill and supply declared or redacted metadata. Keep the complete export for your reviewer.
Claude Code installation and local tool calls were verified. Model-driven skill behavior and Cowork installation need separate checks. This download is a local evaluation package, not a public directory install URL.
Read the install commands and verification limits. A download does not activate a model. Your assistant provider may process conversation and tool content.
Directory status.
Evidence Desk is in testing. Directory status updates here when each listing goes live.
- MCP Registry
- Not listed
- Claude Connectors
- Not listed
- OpenAI directory
- Not listed
- Glama
- Not listed
- Smithery
- Not listed
Open the blank file. Read the processing rules and local setup.
Review boundary.
A review prepares a file. It does not certify compliance, give a legal opinion or connect to a core.
Send the review question.
Tell us the decision, deadline and accountable role. A person will reply within one business day.
Request a call