Skip to content

docs: enterprise review pass on Copilot Agent Standards - #41

Merged
craigthackerx merged 1 commit into
mainfrom
docs/copilot-agent-standards-enterprise-review
Aug 24, 2026
Merged

docs: enterprise review pass on Copilot Agent Standards#41
craigthackerx merged 1 commit into
mainfrom
docs/copilot-agent-standards-enterprise-review

Conversation

@craigthackerx

Copy link
Copy Markdown
Contributor

Review pass on the first edition, against what an enterprise assurance review actually asks for.

Defects fixed

  • Mangled sentence in the declarative agent limits table. The name row read "an agent between the two packages fine and cannot be built in the form as named", which is missing the word it turns on and does not parse.
  • Inconsistent line wrapping. The worked example section was hard wrapped at 100 characters while the other 284 paragraph lines in the document are unwrapped. Unwrapped it.

Gaps closed

The first edition covered how to build an agent well and said nothing about the four things a regulated organisation asks before one ships.

Gap Now covered
Data protection New section: sensitivity labels on grounding data, DLP with the Microsoft 365 Copilot location, oversharing policy for SharePoint and OneDrive, retention, label inheritance into conversations, Insider Risk "Risky AI usage", Communication Compliance jailbreak classifiers, eDiscovery. Ordered per Purview's own four-step deployment model
Tenant controls Copilot Control System settings, the AI Admin role rather than Global Admin, and the point that per-agent hygiene is worthless if anyone in the tenant can publish anything
Third-party agents The external publisher setting is a supply chain decision. Microsoft states plainly that non-Microsoft agent data processing sits outside its agreements
Cost Licensing table: instructions and public grounding are free for everyone, shared tenant data meters in Copilot Credits for unlicensed users, custom engine agents invert both halves, governance carries its own licence floor

Also adds a when an agent misbehaves table. Blocking, disabling the identity, revoking the access package and reverting the artefact are four levers with four different latencies, and an incident is not the moment to work that out.

Anti-patterns (+4), adoption checklist (10 → 14 items), verification table, known gaps and See Also all updated to match.

Verification

  • Five new sources, all fetched live, all returning 200, each recorded in the verification table with Microsoft's own ms.date.
  • No broken internal anchors in the built output.
  • npm run lint, npx tsc --noEmit, npm run build clean.
  • No em dashes or en dashes.

Not done

No diagram. Three of the eight standards docs carry one and five do not, so it is not a house requirement, and the diagrams-as-code pipeline regenerates every SVG in CI. Happy to add one if you want the shipping pipeline drawn.

Two defects and four content gaps found reviewing the first edition
against what an enterprise assurance review actually asks for.

Defects:

- The declarative agent name limit row had a mangled sentence, missing
  the word it turned on, so the row did not say what it meant.
- The worked example section was hard wrapped at 100 characters while
  the rest of the document uses unwrapped paragraphs. Unwrapped it.

Gaps, all of which a bank would raise before a first agent ships:

- Nothing on data protection. Adds a section covering sensitivity
  labels on grounding data, DLP with the Microsoft 365 Copilot
  location, oversharing policy, retention, label inheritance into
  conversations, Insider Risk, Communication Compliance and eDiscovery,
  following Purview's own four-step order.
- Nothing on tenant controls. Adds the Copilot Control System settings,
  the admin roles that reach them, and the point that per-agent hygiene
  is worthless if anyone can publish anything.
- Nothing on agents you did not write. Third-party agent data handling
  sits outside Microsoft's agreements, which makes the publisher
  setting a vendor decision rather than a default.
- Nothing on cost. Adds the licensing table: instructions and public
  grounding are free, shared tenant data meters in Copilot Credits for
  unlicensed users, and governance carries its own licence floor.

Also adds a "when an agent misbehaves" table, since blocking, disabling
the identity, revoking access and reverting the artefact are four levers
with four different latencies and that is not the moment to work it out.

Anti-patterns, adoption checklist, verification table, known gaps and
See Also updated to match. All five new sources verified live with their
own page dates recorded.
@github-actions

Copy link
Copy Markdown

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

@craigthackerx
craigthackerx merged commit dd607ea into main Aug 24, 2026
5 checks passed
@craigthackerx
craigthackerx deleted the docs/copilot-agent-standards-enterprise-review branch August 24, 2026 13:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant