Skip to content

K-Dense-AI/scientific-agent-skills / skills/treatment-plans

treatment-plans

Verified

Formats and structurally validates local treatment-plan documentation after clinical decisions have already been supplied and verified by authorized licensed professionals. Use for source traceability, clinician-authored intervention records, goals and checkpoints, shared-decision records, reconciliation handoffs, and release gates—not for clinical decision-making.

references/documentation_workflow.md

View on GitHub
# Documentation Package Workflow

Last reviewed: **2026-10-01**

## Package contract

A complete package contains exactly one JSON object for each document type:

- `source_fact_manifest`
- `clinician_authored_intervention_record`
- `goals_monitoring_checkpoint_record`
- `informed_preference_shared_decision_record`
- `transition_reconciliation_record`
- `intended_use_handoff_record`

All six must use schema version `2.0` and the same `subject_ref`, `data_classification`, draft status, and notice.

## 1. Intended-use gate

Complete the intended-use/handoff record first:

- purpose and authorized users;
- local setting and accountable roles;
- prohibited-use list;
- privacy and processing attestations;
- local policy, retention, change-control, and reporting routes;
- local emergency-process reference;
- handoff sender, recipient, and acknowledgment;
- clinician sign-off and documentation-handoff release gate.

Do not remove prohibited uses. The generator restores the complete list in every new package.

## 2. Source-fact manifest

Every clinical or process statement used elsewhere must have a source fact. A fact records:

- a stable fact ID;
- fact kind;
- the exact bounded statement supplied;
- source type, title, local locator, and version/date;
- verification status, verifier role, and time;
- applicability status when an official source or policy is involved.

Allowed source categories distinguish signed local records, authorized EHR records, current FDA labeling, current REMS materials, current official guidance, and local policy.

The manifest does not decide which source applies. An authorized professional must confirm applicability.

## 3. Clinician-authored interventions

Each intervention record must:

- state that the decision was supplied and verified by an authorized licensed professional;
- preserve the clinician-authored action without rewriting it into a recommendation;
- link to at least one verified source fact;
- record parameters only as supplied;
- identify the responsible role;
- preserve explicit start/end dates when supplied;
- include verifier role and time.

The validator checks structure and provenance only. It does not parse or judge a medication, procedure, therapy, device, referral, or instruction.

## 4. Goals, monitoring, and checkpoints

Separate:

- goals — statement, measurement, target, target date, and source facts;
- monitoring items — item, method, supplied frequency text, explicit next due date, owner, and source facts;
- checkpoints — exact supplied date, purpose, owner, linked records, and source facts.

Do not convert a narrative frequency into dates. Do not infer a checkpoint from a target, medication, intervention type, standard interval, or prior appointment.

## 5. Informed preferences and shared decisions

Record only what the authorized clinician documented:

- decision topic;
- options actually presented;
- whether benefits, harms, and uncertainty were documented;
- the person's stated preference;
- the clinician-documented outcome;
- participant and author roles;
- source facts, time, and acknowledgment status.

Do not generate missing options, risk estimates, benefit claims, alternatives, or consent language.

## 6. Transition and reconciliation

Record:

- sending and receiving settings and roles;
- exact handoff date;
- medication-reconciliation status;
- source and destination list fact references;
- discrepancy status and authorized reviewer;
- handoff items, owners, recipients, and acknowledgment;
- unresolved items and their local route.

`completed_by_authorized_clinician` is a declaration to be verified against the local record. The script does not perform reconciliation.

## 7. Deterministic checks

Run in this order:

```bash
python3 scripts/validate_treatment_plan.py PACKAGE_DIRECTORY
python3 scripts/validate_traceability.py PACKAGE_DIRECTORY
python3 scripts/check_completeness.py PACKAGE_DIRECTORY
python3 scripts/privacy_process_check.py PACKAGE_DIRECTORY
python3 scripts/check_consistency.py PACKAGE_DIRECTORY
python3 scripts/timeline_generator.py PACKAGE_DIRECTORY --output SCHEDULE.json
```

Run these commands from the skill directory. New empty templates pass structural validation but intentionally fail completeness and schedule generation until supplied records and dates exist.

The checks answer different questions:

- structural validator — are document types, fields, types, enums, bounds, and dates valid?
- traceability validator — do all clinical/process records point to existing verified facts?
- completeness checker — are required records, reviews, routes, acknowledgments, sign-off, and release declarations complete?
- privacy/process checker — are local authorization, minimization, external-tool prohibition, qualified review, retention, and direct-identifier safeguards documented?
- consistency checker — do package IDs, statuses, classifications, references, and explicit date order agree?
- timeline generator — after structural and cross-record consistency checks, what events occur on dates already supplied?

Run every check again after any change. A pass evaluates declarations, not their authenticity. The completeness output is `documentation_declarations_complete`; `handoff_authorized_by_script` is always false. The fixed completeness profile requires at least one intervention, goal, monitoring item, checkpoint, shared-decision entry, and handoff item. These are package constraints, not universal care requirements. If the supplied workflow does not have such records, leave it blocked and retain the authorized source workflow; never invent care or consent documentation merely to pass.

Dates use full `YYYY-MM-DD` form. Timestamps use `YYYY-MM-DDTHH:MM:SS`, optional 1-6 fractional digits, and `Z` or `±HH:MM`. A schedule keeps the calendar date in the supplied offset, without converting timezones. It omits clinical text and subject IDs but is still sensitive when derived from patient data.

## 8. Minimized issue handling

Reports use known field paths, not values; unknown field names and unrecognized document-type values are suppressed. For example, a report may identify `interventions[0].verification.status` without printing the action text.

Resolve each issue in the authoritative local record:

1. Locate the field path.
2. Compare with the signed source.
3. Ask the responsible authorized role to supply or verify the missing value.
4. Update provenance.
5. Re-run all checks.

Never correct clinical content from memory or general guidance.

## 9. Sign-off and handoff

Before releasing for authorized documentation handoff:

- all source facts are verified;
- source applicability is confirmed where required;
- interventions are verified;
- reconciliation is completed or explicitly not applicable by an authorized reviewer;
- unresolved items are routed and acknowledged;
- privacy/process review is complete;
- local policy and reporting routes are populated;
- the recipient is identified;
- the authorized licensed signer completes the attestation;
- the release gate is set to `released_for_authorized_documentation_handoff`;
- blocker codes are empty.

The visible draft/not-medical-advice notice remains. Release does not authorize implementation by an agent.

## 10. Change control

For every revision:

- preserve the prior authorized version according to local records policy;
- update source versions and verification times;
- re-run every check;
- obtain new sign-off when clinical content, recipient, classification, purpose, or governing source changes;
- record disposition of superseded local copies;
- never overwrite a source record or silently reuse an old approval.
rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).Stars: 198kUpdated Claude CodeCodexUnverifiedInstall
doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.Stars: 198kUpdated Claude CodeCodexVerifiedInstall