Compliance review found a mismatch between stated policy intent and implementation for list_repository_collaborators: the PR rationale specified reader integrity, but the guard assigned writer integrity. This change resolves that discrepancy in code and tests so the tool’s label semantics match the documented access-sensitive intent.

• Policy alignment

  • Updated list_repository_collaborators in tool_rules.rs to assign reader_integrity(repo_id, ctx) instead of writer_integrity(repo_id, ctx).
  • Kept secrecy handling as private policy scope for collaborator/permission metadata.

• Documentation-in-code clarification

  • Updated the inline integrity comment from writer-level to reader-level to reflect the intended trust boundary for access-sensitive collaborator data.

• Unit expectation update

  • Adjusted the dedicated test assertion to expect reader-level integrity and updated the assertion message accordingly.

"list_repository_collaborators" => {
    // S = private policy scope
    // I = reader (access-sensitive metadata should not directly authorize writes)
    secrecy = policy_private_scope_label(&owner, &repo, repo_id, ctx);
    integrity = reader_integrity(repo_id, ctx);
}