AI Skill Source Integrations
ConfluenceConfluence

Turn Confluence team knowledge into AI Agent Skills

Develop AI Agent Skills from the reasoning inside runbooks, decision records, and page discussions. Preserve why a rule exists, which exception it allows, and who can explain the boundary.

Keep the explanation in the margin connected to the rule

Source study / Partner operations space

Runbook / Version 6

Partner data handoff

Complete a delivery with a clear reference, a receiving owner, and an explicit acceptance record.

Avery / Process owner
“The reference locates the package. The named owner is the person who can resolve a question about whether it is usable.”
Reasoning to preserve

Keep identity and responsibility together

Retain the source reference and the receiving owner. A folder name alone cannot establish who accepted the work.

Morgan / Operations reviewer
“An upload completed last week, but the receiving team could not use the file. We need their confirmation before closure.”
Reasoning to preserve

Preserve the experience behind the requirement

The added confirmation responds to a failed handoff. Keep that rationale attached so the requirement can be examined in another situation.

Avery / Process owner
“The pilot exception changes the review timing. It does not make an unanswered handoff complete.”
Reasoning to preserve

State what the exception actually changes

The timing allowance has a defined scope. Preserve the acceptance requirement so the exception is not broadened into automatic closure.

A fictional runbook with review context. Select a clause to see the explanation attached to it.

The opportunity

The explanation beside the rule can be the most valuable part

A runbook tells a team what to do. The discussion around it can explain which failure it addresses, what a reviewer challenged, and why an exception is narrower than it first appears.

Develop that reasoning into AI Agent Skills that explain a requirement, recognize its limits, and bring unresolved questions back to the people responsible for the work.

01 / Follow how the guidance changed

Keep the decision trail attached to the current page

The old wording, the reviewer’s explanation, and the recorded decision serve different purposes. Preserve their relationship when developing a candidate skill.

Decision trail / Receiving-owner acceptance

Runbook / Version 5
“Close the handoff after uploading the package.”

The earlier rule made an assumption

The wording treated upload completion as sufficient. Keep this version to understand what the later review changed.

Historical wording / Superseded in this example
Inline comment / Operations review
“The receiving team could not use the file. We need their confirmation.”

The reviewer explains the failure

The comment records why the old condition was insufficient. It is evidence for the reasoning, not a substitute for the agreed requirement.

Review evidence / Explain the reason for the change
Decision D-17 / Runbook version 6
“Retain receiving-owner confirmation. The pilot may use a next-day review window.”

The decision retains the rule and its boundary

The updated runbook requires acceptance and qualifies the pilot timing exception. Keep both connected to the owner and the decision record.

Current example / Acceptance required, pilot timing qualified
A published page and a resolved discussion still need interpretation. The process owner confirms which guidance applies to the next case.

02 / Develop the knowledge into a skill

Make the rule something an agent can reason with

Preserve the requirement, the reason it exists, and the exception’s limits. Test the explanation with the owner before carrying it into another workflow.

Explore skill validation

Start with the experience behind the record

Which rule has a story worth preserving?

Bring the page, the discussion that shaped it, and the person who understands the exception. Explore how that knowledge could become an AI Agent Skill.

Explore your Confluence sources