# Bedside OS — your Chief of Staff for the life around the work

This pack is for staff nurses and APRNs, and it makes one promise with a hard edge: **it carries your life around the work — never the work's clinical core.** Every famous clinical-AI failure happened on the patient side of the line; this module is built to stay on your side of it. Practice-adjacent, never practice-embedded.

And one sentence from the profession's own code, kept where you'll see it: the ANA holds that systems and technologies that assist in clinical practice are *adjunct to, not replacements for, the nurse's knowledge and skill* — nurses remain accountable for their practice even when the technology fails. This tool never shares your accountability. That's not a limitation; it's the design.

**What's inside**

| File | What it does | Tier |
|---|---|---|
| `Shift-Rhythm.SKILL.md` | Pre-shift brief and post-shift decompression, on your own schedule | 🟢 Green |
| `CE-Tracker.SKILL.md` | License, certification, and CE timelines that never sneak up on you | 🟢 Green |
| `Portfolio-Builder.SKILL.md` | Clinical-ladder, scholarship, and BSN→DNP application drafts — approved by you | 🟡 Yellow |
| `Case-Study-Scrubber.SKILL.md` | Teaching cases with the de-identification gate *before* drafting begins | 🟡 Yellow |
| `Evidence-Reader.SKILL.md` | Rubric-driven paper summaries under the research gate | 🟢 Green |
| `Judgment-First.SKILL.md` | The anti-anchoring pattern: your assessment first, the AI's draft second | 🟢 Green |
| `Bedside-Boundary-Header.md` | The shared boundary all six skills load — the strictest in the OS | — |
| `After-the-Shift.md` | The decompression ritual: leave the shift at the door without losing its lessons | — |

**Setup (2 minutes).** Tell your AI: *"Load the six skills in 18-Bedside-OS, including the Bedside-Boundary-Header they share. Apply my Do-Not-Remember rules with the Bedside/APRN extension — the strict one. My shift pattern is [days/nights/rotation]; schedule Shift-Rhythm around it."*

**The rules that make this safe at the bedside.**
- **Nothing patient-adjacent, ever** (🔴): no patient-specific questions, no chart content, no story-shaped fragments that could re-identify ("the 34-weeker from Tuesday" counts). Refused at intake, not stored-then-scrubbed.
- **Your judgment first.** Anything that even resembles decision support triggers the Judgment-First pattern: you state your own read before the AI shows you anything.
- **Employer systems are off-limits** without formal approval: no EHR, no scheduling system, no employer Teams/Slack (🟠 deferred → effectively 🔴 unapproved). Connect your *personal* calendar, not the unit's.
- External-facing drafts (applications, emails, anything submitted) are 🟡 — you approve the exact text.
- If a session turns to crisis territory — yours or anyone's — the task stops and human resources come first (🔴).

**Worth tracking** (adoption signals only; never patient-outcome or performance claims): time reclaimed by your own estimate, CE progress, whether the decompression ritual actually gets used, wellbeing pulse — self-reported and separate from everything else.

> Agents propose. Humans judge. Nurses steward. · If it could identify a patient, it counts.