@maverickcer/env-cap
    Preparing search index...

    Interface VariableEvidenceDocs

    Developer-supplied evidence assertions for one variable -- categorically different from every field on VariableDocs: those are declared-and-never-verified, while dynamicAccess is re-checked against reality on every run (fresh / stale / missing). Kept structurally separate, in VariableDocs.evidence rather than folded in alongside description/ owner/..., specifically so that different epistemic status is visible in the shape itself and not only in a doc comment. See ADR 0037.

    interface VariableEvidenceDocs {
        dynamicAccess?: readonly string[];
    }
    Index
    dynamicAccess?: readonly string[]

    Citation(s) of where this variable is actually read dynamically -- somewhere env-cap's own static AST scan can't see (a shell script, a Docker entrypoint, a sibling service). Each entry is a "<relative-path>:<line>:<column>" citation.

    A developer's re-acknowledgment that access happens, never a claim env-cap itself observed anything -- tracked as its own independent fact and never folded into the AST-derived VariableAccessStatus. Re-verified every run: a citation whose file no longer exists, or whose content has visibly changed since it was last acknowledged, is flagged rather than trusted forever.