From gxp-review-general
Conduct a structured Quality or Technical review of any GxP document or record, classify findings as Critical/Major/Minor/Observation against FDA, EU, ICH, and applicable Montai SOPs, and deliver a Word report. Use whenever someone asks for a "GxP review", "QA review", "quality review", "technical review", "SME review", "draft review", "pre-approval review", "post-approval review", or "compliance check" on any controlled document (SOP, policy, protocol, plan, specification, method, master record) or GxP record (batch record, log, certificate, training record, deviation, CAPA, change control, audit report, validation report). This is the GENERAL-PURPOSE GxP review skill — use it for Montai-internal docs, vendor docs outside the CMO pre-approval playbooks, drafts being authored, and post-approval retrospective audits.
How this skill is triggered — by the user, by Claude, or both
Slash command
/gxp-review-general:gxp-review-generalThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill produces a structured, regulator-defensible review of any GxP document or record. The deliverable is a Word review report with findings classified as **Critical / Major / Minor / Observation** and a status-appropriate disposition.
This skill produces a structured, regulator-defensible review of any GxP document or record. The deliverable is a Word review report with findings classified as Critical / Major / Minor / Observation and a status-appropriate disposition.
The skill works for:
Before any review begins, the skill needs three pieces of information. If any one is unclear from the request and uploaded materials, ask the user up front using a single combined ask_user_input_v0 call — do not start reviewing.
| Question | Options |
|---|---|
| Reviewer role | Quality (QA / QC / Compliance) · Technical / SME · Combined (one person playing both) |
| Document status | Early Draft (in authoring) · Final Review (pre-approval / pre-signature) · Post-Approval (retrospective / audit-style) |
| GxP discipline | GMP · GCP · GLP · GDP · GVP · GAMP 5 / Part 11 / Annex 11 (computerized systems & data integrity) · Other (ask which) |
Inferring is fine when context is clear:
Optional follow-up (ask only if material to the review): regulatory jurisdiction (FDA only / EU only / both / ICH-aligned global default). When unstated, default to both FDA and EU with ICH harmonization — Montai's pipeline is U.S.-IND-led but EU expansion is on the horizon.
Once the three answers are in hand, restate them in one line ("Quality reviewer · Final Review · GMP — proceeding") and continue.
Choose the tool by source:
app.box.com, *.box.com) → Box:get_file_content for text extraction; Box:get_file_preview for visual inspection of PDFs/images. For folders, Box:list_folder_content_by_folder_id.montai.sharepoint.com) → Microsoft 365:read_resource with the file URI; resolve via Microsoft 365:sharepoint_search if needed./mnt/user-data/uploads/ → for .docx, run pandoc --track-changes=all input.docx -o output.md to capture tracked changes; for complex .docx with forms / embedded objects, also unpack via python /mnt/skills/public/docx/scripts/office/unpack.py to inspect the underlying XML. For .pdf, read /mnt/skills/public/pdf-reading/SKILL.md first.Whatever the source, scan for these red flags during retrieval (they often become findings on their own):
Error! Reference source not found, {错误!未找到引用源})TBD, <XX>, XXXX, [Insert …])Read references/document-type-checklists.md. It covers:
Each section contains a focused pass/fail checklist. If the document spans types (e.g., a method that includes a validation report), load both checklists.
Always load references/severity-and-regulatory.md as well — it contains severity definitions, the FDA / EU / ICH / USP / OECD / PIC/S regulatory framework, and how to choose citations.
Read references/reviewer-perspectives.md for the Quality vs. Technical vs. Combined focus and references/status-modulation.md for how the document status changes review depth and severity grading.
The short version:
| Early Draft | Final Review | Post-Approval | |
|---|---|---|---|
| Quality reviewer focus | Template / structure conformance, regulatory framework alignment, gap-finding | Full compliance, signatures, traceability, ALCOA+, training linkage, ready-to-sign call | Audit-perspective: change-control / CAPA triggers, deviation linkage, periodic-review readiness |
| Technical reviewer focus | Scientific logic, methodology soundness, feasibility, calculations | Verify all technical content; reproducibility; specification linkage | Verify the document still reflects current technical reality; flag stale procedures |
| Severity bias | Cap most findings at Major; reserve Critical for fundamental flaws | Full C/M/m/O range applies | Critical findings here typically trigger immediate change control / quarantine — be deliberate |
This framing is captured in the report's "Review Scope" section so the reader knows which lens was applied.
Apply the loaded checklist in roughly this order:
references/severity-and-regulatory.md for the citation menu.references/montai-sop-policy-index.md to identify which Montai POL/SOP/FRM apply, and fetch them on demand from Box folder 330567208254 when their content is material to a finding (e.g., a CAPA write-up requires checking SOP-04; a batch release record requires checking SOP-06 and FRM-06-01). Do not bulk-fetch all SOPs at the start of the review — fetch only what's relevant.Every finding has six required elements:
F-{NN} (e.g., F-01, F-02) within the report; if the user maintains a Findings Register, propose register-style IDs (e.g., {DocID}-C01, {DocID}-M03) on hand-off.references/severity-and-regulatory.md.Order findings: Critical → Major → Minor → Observation, then by document section within each severity tier.
Before rendering the report, re-read the source and verify each finding:
If a finding fails self-check, either correct it against the actual source or remove it. Better to issue 8 solid findings than 12 where one is fabricated. A single hallucinated finding damages reviewer credibility and delays approval cycles.
When complete, state in chat: "Self-check complete: N findings verified against source."
Use the docx skill at /mnt/skills/public/docx/SKILL.md. The output skeleton is in assets/review-report-template.md — follow it.
Required sections (all):
references/severity-and-regulatory.md| Status | Disposition options |
|---|---|
| Early Draft | Continue authoring with revisions · Substantive rework required · Restructure (template / approach mismatch) |
| Final Review | Approve as written · Approve with editorial revisions · Conditional approval pending Major resolution · HOLD — do not approve (Critical or unresolved Major) · Reject |
| Post-Approval | No action required · Initiate change control to address findings · CAPA required (per SOP-04) · Supersede / revise document · Withdraw document |
Pick exactly one. The disposition should follow automatically from the findings — a single unresolved Critical at Final Review = HOLD; multiple unresolved Majors = HOLD or Conditional.
File naming: {DocID}_GxP-Review_{YYYYMMDD}.docx, saved to /mnt/user-data/outputs/. If the document has no ID, use a slugged version of the title.
Use present_files to deliver the Word report. In chat, include only:
Do not restate the full report in chat. The Word file is the deliverable.
MR-0033809, MTAI-1025); afterward, the short form is fine.2026-04-28) or 28 Apr 2026; never US-style numeric (04/28/2026) in regulatory text — it's ambiguous to non-US readers.A review is acceptable for delivery when:
/mnt/user-data/outputs/ and was delivered via present_filesCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub wshayes/skill-marketplace --plugin gxp-review-general