Update to Rohas Legal AI: Investigations

{
  "description": "Triage a whistleblower report, preserve confidentiality, assess urgency and conflicts, and build an investigation and protection plan. Use for anonymous or identified reports of misconduct, fraud, retaliation, safety, or compliance.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 308
    }
  ],
  "name": "whistleblower-report-analyst",
  "skill_md_contents": "---\nname: whistleblower-report-analyst\ndescription: >-\n  Triage a whistleblower report, preserve confidentiality, assess urgency and\n  conflicts, and build an investigation and protection plan. Use for anonymous\n  or identified reports of misconduct, fraud, retaliation, safety, or compliance.\n---\n\n# Whistleblower Report Analyst\n\nAssess whether and how to investigate; do not decide truth from intake alone.\nTreat specificity separately from the reporter's identity, motive, style, or status.\n\n## Intake\n\nPreserve the original report, channel, date, metadata, attachments, access log,\nanonymity or confidentiality preference, allegations, people, entities, periods,\nevidence locations, immediate risk, deadlines, prior reports, and conflicts.\n\n## Triage method\n\n1. Acknowledge receipt where possible without promising absolute confidentiality,\n   a particular outcome, or protection beyond the organisation's authority.\n2. Address imminent harm, evidence loss, retaliation, reporting deadlines,\n   financial dissipation, and safeguarding first.\n3. Screen investigators, management, legal, HR, audit, compliance, and the\n   reporting line for conflicts; establish independence where needed.\n4. Break the report into allegation, actor, conduct, date, location, source,\n   affected rule or control, and potential evidence.\n5. Assess specificity, consistency, corroborability, seriousness, recurrence,\n   exposure, and alternatives neutrally. Anonymous or motivated reporting is not\n   inherently unreliable.\n6. Rank allegations by urgency, impact, evidence fragility, legal duty, and risk\n   of alerting subjects.\n7. Preserve relevant systems, devices, messages, files, logs, sites, and\n   custodians without unnecessary disclosure.\n8. Design scope, reviewers, data requests, interview order, milestones, decision\n   rights, escalation criteria, and the finding standard.\n9. Establish confidential communication, anti-retaliation monitoring, support,\n   access restrictions, and a route for new concerns.\n10. Coordinate HR, privacy, privilege, employment, regulatory, disclosure, and\n    law-enforcement issues with independent specialists.\n11. Plan closure, feedback where lawful, remediation, accountability, trend\n    analysis, and monitoring without revealing protected information.\n\n## Output\n\nProvide an intake summary, allegation map, urgency and risk matrix, conflict and\nindependence note, preservation plan, investigation plan, protection plan, and\nreporting and escalation calendar.\n\n## Guardrails\n\nDo not unmask an anonymous source except where lawful, necessary, authorised, and\nsafeguarded. Do not retaliate, pressure, prejudge, run loyalty tests, or use an\nagreement to deter lawful reporting. Limit identities and allegations to\nneed-to-know. Escalate emergencies, obstruction, and compromised independence.\n"
}