When to use it

Use independent-research when a claim needs checking against named local files or repositories and you want the evidence gathered in a separate worker context. Frame a concrete question and identify the roots it is allowed to inspect.

How it works

Figure 01

Keep the investigation bounded and traceable.

  1. Scope

    Name the question and exact authorized local roots.

  2. Inspect

    Use a fresh read-only worker and a fixed source budget.

  3. Return

    Report evidence, provenance, and access gaps.

ResultAn evidence packet for the calling agent
The worker gathers evidence; the caller owns the conclusion. If the required context isolation is unavailable, the skill reports a gap.

The skill uses a bounded read-only worker to inspect local sources and return an evidence packet. The current collection budget allows up to eight source artifacts and 64 KiB, keeping the result focused enough to examine.

Source digests and no-follow reads help identify what was actually inspected. The calling agent uses that packet to assess the original question while retaining responsibility for the conclusion.

An example request

Ask your agent

“Independently check whether the migration runner can apply the same migration twice. The authorized root is ./ (the current repository); inspect its migration code and tests.”

This illustrates a request you can adapt; it is not a transcript of a completed run.

An evidence packet tied to the inspected source, with enough provenance to revisit the claim and see what the bounded investigation covered.

Install and use

The skill’s identifier is independent-research. It ships in the critical-thinking plugin. In Claude Code, first add the Overclock marketplace, then install the package:

/plugin install critical-thinking@overclock

For a standalone installation, keep the skill directory and its bundled resources together. The repository also includes per-skill Codex metadata. Read the full skill instructions for its exact workflow and supporting files.

What to keep in mind

The worker’s scope and read-only contract are not a claim of an operating-system sandbox. If the required isolation is unavailable, the skill reports that limitation rather than presenting the work as independent.

This is a local-source investigation, not open-ended web research. A bounded sample can leave relevant evidence outside its coverage.

  • Critical thinking — Test a proposal’s assumptions before accepting its conclusion.
  • Agent bridge — Ask another installed coding agent for a review or a bounded implementation.