Diligent is a portable Agent Skill for auditing U.S. government websites and digital services. Every step leaves a durable artifact, and the final status is computed, not asserted.
npx skills@latest add Simple-Cortex/diligentWith no failure and nothing unresolved, the status is pass_for_declared_scope, for the declared scope only.
Only because nothing was left untested.
pass_for_declared_scope. With no failure and nothing unresolved, the status is pass_for_declared_scope, for the declared scope only. Only because nothing was left untested. Exit code 0.
Diligent 0.1.0 runs no automated website checks. Results come only from human or documentary input.
A small model of the gate's own rules, run in your browser. Switch a condition on to see the outcome it produces.
Version 0.1.0 is a contracts-and-gate release. It ships zero automated website checks. It defines the workflow, the artifact formats, the authority registry, the rule packs, and a deterministic gate. It does not find or clear accessibility issues by itself.
| Observed, and it meets the rule | |
| Observed, and it does not | |
| Routed to a person to decide | |
| The check did not run | |
| The check could not run | |
| The check gave no usable result |
Every result is pass, fail, needs_review, not_tested, blocked, or error. A skipped test is not_tested, never a pass.
Read the result statesAn unknown applicability keeps the assessment incomplete, and incomplete can never become a pass. The gate exits 2.
See what each mode can claimverify runs in a context that did not write the repair, from frozen evidence. A pull request or a green build is not proof that a barrier is gone.
Read the non-negotiablesThe implementation is zero-dependency Node ESM, plain JSON and Markdown, with no framework and no build step. Everything an agent needs to act on is a file it can read.
See the install stepsArtifacts live under .diligent/<audit-id>/. Each phase writes its artifact before the next may start.
The CLI runs init, status, gate, report, and rules validate in 0.1.0; every other phase is agent-performed.
You need the audit to say what it covered and what it did not. Diligent reports counts with denominators, lists unsupported criteria, and routes judgment calls to named roles.
You need to know where to fix something and what must not change. Findings carry a location, a reproduction, an acceptance test, and the protected constraints.
You need the right questions at the right time. Evidence requests name the exact artifact, the owner role, and the obligation it would settle.
Requires Node 20 or later for the command line helpers. The skill also has a manual fallback for hosts without scripts.
npx skills@latest add Simple-Cortex/diligentAlso from the same maker: Considered, the design-reasoning skill.