interrogate

---

name: interrogate description: Plan and execute an evidence-first interrogation using approved probes; outputs schema-valid findings + report. metadata: source_repo: https://github.com/jscraik/Agent-Skills source_rev: 7e31061c353c94746910d239ae122900cc5324fb-dirty source_dirty: 'true' source_dirty_paths: utilities/recon-workbench/references/evals.yaml, utilities/skill-creator/scripts/run_skill_evals.py, design/better-icons/

You are operating in a reverse-engineering inspection context.

Guardrails:

• Only analyze targets the user owns/controls/has permission to test, or open-source repos.
• Do not propose or run DRM/TPM bypass, decryption/circumvention, or credential theft.
• Use only probes listed in probes/catalog.json unless the user explicitly expands the catalog.
• Every claim must cite an artifact path (runs/.../raw/...).

Workflow:

1. Run meta.doctor if toolchain readiness is unknown.
2. Produce a ProbePlan that includes:
   • baseline run
   • at least one stimulus run if goal involves behavior
3. After probes execute, summarize into findings.json per schemas/findings.schema.json.
4. If static signal is low or protections are detected, switch to $worst_case_interrogation.

Output:

• When planning: JSON matching schemas/probe-plan.schema.json
• When summarizing: JSON matching schemas/findings.schema.json