Skip to content
Documentation

Evidence (beta)

What it does

Evidence reads the systems an application runs on and records what it found against its directives, for a person to decide on. Four rules hold:

  • Checks only read. Only operations cajeX has classified as read, and cajeX asks each platform for read access only.
  • The same steps, every run. A person approves a check's steps once; every run makes exactly those calls.
  • A person records. cajeX proposes a result with its reasons and the data it read. A fail always waits for a person.
  • Evidence expires. When a result's validity window closes, the check reads Stale, not Passing.

How to use it

Evidence in the left menu shows each application's setup and what waits for a person, and guides a first run. Each application's Evidence tab is where the work is done: a setup checklist, Needs you, and the record.

Setting up an application

  1. Connect a platform under Admin → Connections (a workspace admin, once): the cajeX Evidence app or a fine-grained token for GitHub, a service account for Jira and Confluence. An application's team can add access for its own repository, or ask the platform team.
  2. Say where it runs. Link the application to its repositories and projects; cajeX suggests matches. A check reads only what is linked.
  3. Link the directives that govern it, on its Conformance tab.
  4. Set up evidence. From its catalogue, with no AI model, cajeX works out how each directive can be shown: which ready-made checks fit, what needs a repository first, and what only a person can attest. This is the evidence plan.
  5. Dry run, then approve. A dry run reads for real and records nothing. One approval switches the checks on; nothing runs before that.

Four ways a directive can hold

Kept separate, as they differ in strength: Decided (an accepted decision record says so), Reviewed (documents, the design or a code review show it), Proven (a check read a system and its result was recorded), Attested (a person confirmed it and said what they looked at).

Collecting and results

Collect evidence runs an application's checks now. Collection sets what happens otherwise: Off (only when someone collects), Shadow (on a schedule; every result waits for a person) or Assisted (on a schedule; a pass decided by an approved check records itself, a fail always waits, and a weekly sample of self-recorded passes comes back to a person). With the cajeX Evidence app and On change, a push to the default branch, an alert (code scanning, secret scanning, Dependabot) or a release re-runs the checks that read it; ruleset changes wait for the next scheduled run.

A result reads Passing, Failing, Awaiting review (waiting for a person), Couldn't check (cajeX could not read enough to decide, usually access; the check's steps say what was missing) or Stale. A failing control raises one finding listing every directive it affects.

Attestations, the trail and auditors

Record attestation is for what no system can show: say what you looked at and attach files. Where your workspace asks for it (Admin → Connections), a second person who is not the attester confirms. Cited files cannot be replaced or removed.

A directive's evidence trail lists every result, attestation and correction in order. The record is only added to, and each day is sealed; Verify trail names any entry that no longer matches. Export evidence package gives an auditor a period's evidence with a verifier that needs no access to cajeX. Full outputs and files are kept for your plan's retention period; a legal hold stops their deletion.

Also

  • On a pull request: where a workspace admin turns it on, cajeX posts a check run saying which of the design's dependency rules a change would break, using the separate cajeX Evidence Writer app. Shadow records what would be posted and sends nothing.
  • On the diagram: pin a linked repository to a shape to show its checks' state, on a layer you can hide.
  • A ticket for a finding: Raise a Jira issue from a finding, linked back to it. A workspace admin turns it on per application (Off, Shadow or Live, and the project) and adds a separate writer credential under Admin → Connections → Jira → Let cajeX write.

Availability

Evidence is in beta, available on request: ask cajeX to turn it on for your workspace. It needs Applications on. Its connectors, GitHub, Jira and Confluence, are in beta too.

Permissions

A workspace admin connects platforms and turns on anything cajeX posts to another system. An application's managers, or a workspace admin, approve its evidence plan. Anyone who can see the application can see its evidence and record a result; a confidential application's evidence is visible only to its team.

Tips

  • Start with one application on Shadow; read a few weeks of results before letting passes record themselves.
  • When two connections read the same repository, cajeX marks one as duplicate access; removing the narrower one loses no check.