Skip to content

Further downrank docs/tests/config findings by default #5

@EgemennSahin

Description

@EgemennSahin

Problem

The agent repo lab found that docs/tests/config-like files can still appear with risk wording. They should remain visible for auditability, but they should not create action-required pressure unless the repo explicitly configures them as runtime prompt, scenario, or agent input.

Evidence from tested repos

From .agentdiff/agent-repo-lab/latest/report.md:

  • vercel-labs/lead-agent: README.md was flagged as agent_entrypoint/docs_example with state_mutation and external_side_effect risk wording.

This is useful as a noise signal for agentdiff. It is not a claim that the external repo has a bug or vulnerability.

Proposed fix

Further downrank docs/tests/config findings by default:

  • Keep them visible in reports.
  • Prefer informational language for docs/examples/tests/config unless reachable from runtime agent code or explicitly configured.
  • Preserve suppression suggestions with reason and expires.
  • Avoid action-required severity for README/docs-only agent vocabulary unless stronger runtime evidence exists.

Acceptance criteria

  • npm test passes.
  • npm run lab:agent-repos still has 0 crashes.
  • The cited vercel-labs/lead-agent README-style finding is lower pressure or clearly informational.
  • Suppressed/low-confidence findings remain visible; they do not disappear silently.
  • Runtime reachable high-risk tools are not downranked accidentally.

Non-goals

  • Do not delete docs/test/config findings entirely.
  • Do not add a complex suppression expression language.
  • Do not accuse external repos of unsafe behavior.
  • Do not require installs, API keys, or live model calls.

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentationgood first issueGood for newcomers

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions