Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add vinayaklatthe/microsoft-security-skills --skill purview-data-classificationgit clone --depth 1 https://github.com/vinayaklatthe/microsoft-security-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/vinayaklatthe/microsoft-security-skills/purview-data-classification)<a href="https://agentmods.dev/skills/vinayaklatthe/microsoft-security-skills/purview-data-classification"><img src="https://agentmods.dev/badge/skills/vinayaklatthe/microsoft-security-skills/purview-data-classification/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/vinayaklatthe/microsoft-security-skills/purview-data-classification"><img src="https://agentmods.dev/badge/skills/vinayaklatthe/microsoft-security-skills/purview-data-classification.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00106 | $0.01259 |
| Opus 5.5 | $0.00042 | $0.00504 |
| Sonnet 5.5 | $0.00021 | $0.00252 |
| Haiku 4.5 | $0.00011 | $0.00126 |
Grade A, and why
purview-data-classification scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Microsoft Purview Data Classification & Sensitivity Labels
Classification identifies sensitive content; sensitivity labels apply persistent protection (visual marking, encryption, access, and downstream DLP/lifecycle controls). Together they are the foundation of Microsoft Information Protection and every Purview enforcement story.
When to use
Defining what data is sensitive and applying a consistent label taxonomy across Microsoft 365, Power BI/Fabric, and (via Purview Data Map) the broader estate.
Do not use this skill for AI prompt visibility (use purview-dspm-ai) or for full strategy
sequencing (use purview-information-governance).
Pick the right detection method
| Use case | Method |
|---|---|
| Well-known regulated patterns (credit card, NI, SSN, IBAN) | Sensitive Information Types (SITs) - built-in |
| Org-specific keywords / regex | Custom SITs with confidence tuning |
| Concept-based ("this looks like a contract") | Trainable classifiers (built-in or custom) |
| Match against your own customer/HR list | Exact Data Match (EDM) - hashed structured data |
| Hybrid - regex + supporting evidence | SIT with named entities + proximity |
Rule of thumb: SITs for patterns, classifiers for concepts, EDM for "is this row from our actual customer table". Don't try to make one method do all three jobs.
Approach
- Inventory regulatory and business drivers - List the data types you must detect (GDPR PII, PCI, IP, HR) and map each to a detection method. Verify: each detection candidate has a named business owner.
- Choose detection methods - Pick SITs, trainable classifiers, or EDM per data type; build or tune confidence levels. Verify: a test corpus of true and false positives is prepared per detector.
- Design a label taxonomy - Keep it simple (e.g., Public / General / Confidential / Highly Confidential) with sub-labels; define markings, encryption, and scope per label. Verify: ≤4 top-level labels, ≤3 sub-labels each, named with business language not jargon.
- Publish labels - Use label policies to scope labels to users; set defaults and mandatory labelling where appropriate; configure mandatory-label justification. Verify: pilot users see the label menu in Word/Outlook with the correct default.
- Auto-labelling - Use client-side (recommended/automatic in apps) and service-side auto-labelling (SharePoint/OneDrive/Exchange) based on SITs/classifiers; always start in simulation. Verify: simulation reports show match counts per label and per source; review false-positive sample.
- Roll out in waves - Pilot department -> business unit -> tenant; pair every wave with training and a feedback channel. Verify: label adoption metric per wave > target before promoting.
- Extend - Labels flow into DLP, Data Lifecycle, DSPM for AI, and Defender for Cloud Apps - confirm downstream policies key off the new labels.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 90 lines · 106 tokens per session scan A 471388336c8a
purview-data-classification is a skill published in the GitHub repository vinayaklatthe/microsoft-security-skills (178 stars, last pushed 4mo ago), licensed MIT. It adds 106 tokens to every session and 1,259 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
power-automate
Use when operating Microsoft Power Automate cloud flows from code — create, enable, update, list or delete via the Dataverse Web API (workflow table, category 5) with Entra ID OAuth2, plus run-history debugging. NOT designing the flow definition (that is automation-flows), NOT picking a platform by billing model (that…
rfp-response-compliance-review
Reviews a draft proposal against the firm's configurable compliance rules — prohibited language, required disclosures, role and positioning boundaries, and procurement-submission requirements. Logs all findings to the Review Findings SharePoint list and annotates the draft document for salesperson and reviewer…
carta-witness-signatures
Status of witness and spousal-consent signature requests on option grants, RSAs, PIUs — who still needs to sign, and what's awaiting signature, signed, or expired. Covers one award, one company, or a whole portfolio. Read-only.
carta-hmrc-limits
Report a UK company's HMRC EMI/CSOP limit usage. Use when asked how much EMI allowance is left, whether the company is near the £6m EMI limit, who is close to or over their £250k individual limit, CSOP headroom, or whether a planned grant fits.
patentforge-provisional
Use this skill when the user wants to turn an invention idea into a structured invention disclosure, prior art search strategy, claim candidates, and provisional patent draft package. This skill is for patent drafting assistance only and must not provide legal advice or guarantee patentability.
carta-interaction-reference-core
Retired. Do not load. Carta's answer rules are in carta-cap-table:carta-cap-table.