メニュー
Use for digital organization OrgOps audits, stale lease scans, orphan detection, blocker pattern analysis, memory hygiene, skill/template improvements, milestone retrospectives, and system health reports.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when reviewing a digital organization worker output for scope match, defects, assumptions, domain quality, forbidden actions, evidence gaps, and readiness for QA or verification.
Use for digital organization verification roles: acceptance criteria mapping, evidence checks, UAT readiness, Done transitions, and rejection or blocker routing.
Use when executing one leased task from a digital organization task pool; covers scoped work, Paperclip issue/run state, authority checks, evidence, blockers, heartbeat, and handoff to reviewer/verifier/orchestrator.
Use when managing a digital organization task pool across Paperclip-owned goals, projects, workstreams, stages, leases, blockers, worker/reviewer/verifier assignment, digests, kill switch, UAT, and archival.
Use when activating the digital-org orchestration model in any new project, including Paperclip-native setup, onboarding, discovery, research, kickoff, source-of-truth map, authority profile, capacity policy, and first task slices.
Use as the development domain pack for digital organization tasks involving code, repositories, tests, builds, runtime checks, deploys, DB migrations, code review, QA, and developer evidence.
| name | digital-org-auditor |
| description | Use for digital organization OrgOps audits, stale lease scans, orphan detection, blocker pattern analysis, memory hygiene, skill/template improvements, milestone retrospectives, and system health reports. |
watchdog_mode:
use_when:
- active_work_is_running
- active_tasks_may_lack_workers
- active_workers_may_lack_valid_tasks
- leases_may_be_stale
- work_may_be_looping_or_drifting_from_goal
checks:
- active_paperclip_issue_has_checkout_or_live_run_or_orphan_marker
- active_worker_agent_run_has_valid_issue
- role_agents_have_readable_project_task_role_names
- provider_session_id_is_not_used_as_durable_role_owner
- active_lease_holder_matches_paperclip_checkout_or_live_run
- issue_checkout_live_run_agent_registry_and_control_record_agree
- heartbeat_due_at_not_overdue
- material_delta_since_last_heartbeat
- user_questions_are_in_Questions_with_question_first
- peer_runtime_or_stale_blockers_are_not_misrouted_to_Questions
allowed_actions:
- mark_orphan_needs_worker
- mark_lease_stale
- mark_blocked_stale_or_looping
- request_worker_handoff
- escalate_to_project_orchestrator
- write_compact_watchdog_digest
Use this for organization health, not ordinary task execution.
Audit the system around the work:
Questions polluted by peer/runtime/stale blockers;Deferred issues with active leases;Reporting or Done before UAT authority is recorded;Memory should become reusable operational knowledge:
Do not write noisy memory after every small task. Prefer milestone or audit updates.
Record tool degradation instead of hiding it. If a watchdog or auditor cannot use a helper such as context-mode, it should continue with bounded direct reads when safe, write the degraded state, and hand off. If any role loops without material delta, audit whether it wrote a current artifact or explicit gap before stopping.
System health:
Top bottlenecks:
Orphaned work:
Agent/run identity issues:
Issue checkout/live-run/control divergence:
Repeated blockers:
Weak evidence patterns:
Authority risks:
Memory cleanup/update candidates:
Skill/template improvements:
Recommended next actions:
When kill switch is active, verify no new leases are issued, active workers are stopped or marked orphaned, runtime/public actions are forbidden, emergency digest exists, and recovery options are explicit.