The skills in this directory target the Atlassian Rovo MCP v2 endpoint,
https://mcp.atlassian.com/v2/mcp. They are what every packaged plugin installs, and each one
lives at skills/<name>/SKILL.md per the Agent Skills convention.
These workflows answer in the conversation first. They offer an HTML report when requested, and check whether Atlassian Artifacts is available before offering publication. Results depend on the user’s permissions, available Atlassian apps, and connected sources.
| Skill | Use it to |
|---|---|
assess-release |
Assess release readiness, blockers, and outstanding conditions. |
before-you-start |
Find similar efforts and their recorded owners before starting work. |
catch-up |
Review decisions, changes, and work needing attention during a specified period. |
decision-trace |
Trace a decision, its alternatives, and later changes to source records. |
dependency-map |
Map dependencies, their direction, and blocked or at-risk work. |
how-it-works |
Explain product behavior from connected source code and supporting records. |
The workflows share teamwork-graph-context for context
gathering and atlassian-design for optional report exports.
Both dependencies are included in the package, along with the report renderers, examples,
and design assets. The included live-artifact adapter applies only to its named host;
other hosts use their documented capabilities or the static report fallback.
The previous generation is archived in v1/, unchanged, for anyone still pointed at a v1
endpoint (/v1/mcp/authv2 or /v1/mcp). No plugin installs the archived skills — plugin
discovery only looks one level deep for a SKILL.md, so v1/ is skipped. Set them up manually if
you need them.
A skill written for one generation will fail against the other. The most consequential differences:
discover and execute meta-tools. Each
skill that needs one explains the convention in its Calling non-primary tools section.*Page to *Content — getConfluencePage →
getConfluenceContent, createConfluencePage → createConfluenceContent,
updateConfluencePage → updateConfluenceContent, and searchConfluenceUsingCql →
searchConfluence. The v1 names survive only as search-recall synonyms, not callable operations.detail="full" for the body; document edits require a snapshotToken from the preceding read;
writes must honor space instructions via getConfluenceSpace; and create/update take a parent
object plus a required contentType, with body as {format, value}.getVisibleJiraProjects → listJiraProjects,
getJiraProjectIssueTypesMetadata → listJiraProjectIssueTypesMetadata, addCommentToJiraIssue → addOrEditJiraIssueComment.createJiraIssue takes issueType and assignee, not
issueTypeName and assignee_account_id.For the current tool reference, see Supported tools.