메뉴
Use when the user wants to connect to Azure DevOps — query Boards work items with WIQL, read PRs, inspect pipeline runs, create or update work items and wiki pages, or set up ADO auth. The connector — not security tradecraft.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | azure-devops |
| description | Use when the user wants to connect to Azure DevOps — query Boards work items with WIQL, read PRs, inspect pipeline runs, create or update work items and wiki pages, or set up ADO auth. The connector — not security tradecraft. |
Connector for Azure DevOps (Boards, Repos, Pipelines, Wiki, Test Plans) via Microsoft's official MCP server. The agent sees the enabled tool surface (~76 tools on the shipped domains) once it loads — this skill covers auth, the domain / toolset model, WIQL, and the idioms. Security triage methodology — dedup, severity, SLA, owner inference — is deliberately out of scope; pair this with whichever triage capability your workflow uses.
The manifest pins --authentication azcli, so the MCP delegates to your existing
Azure CLI session — no token to paste, no per-call browser prompt.
az login with an account that can reach the target ADO org.ADO_ORG (the org name, e.g. contoso, not the URL) in the secrets
screen (/secrets or F7) or your launching shell. Optionally ADO_PROJECT /
ADO_TEAM — when a tool needs a project or team and the call didn't name one,
the server elicits a picker; these answer it up front. Needs the pinned
@azure-devops/mcp ≥ 2.8.0; they are inert on older builds.checks: fail fast if az is missing or az account show is empty — by
design, since azcli has nothing to delegate to without a live session.First tool call may show a one-time org-consent browser prompt — that's Microsoft's MSAL flow, not ours.
To switch modes, edit --authentication in capability.yaml:
| Mode | What | When |
|---|---|---|
azcli (default) | Reuses the az login token | Box already signed into Azure CLI |
interactive | Browser sign-in to Entra | No CLI session; human present |
pat | base64 PAT from PERSONAL_ACCESS_TOKEN | Headless / non-Entra orgs |
envvar | Raw bearer from ADO_MCP_AUTH_TOKEN | Token minted elsewhere |
Read vs write is enforced by project roles — a read-only principal can't mutate;
no in-band toggle. (The remote preview at https://mcp.dev.azure.com/{org} is
Entra-OAuth-only and doesn't support Claude Code yet — the local server is the right
default.)
The upstream defaults to all (90 tools); the capability narrows that to 76 via
-d. Add domains as separate list items in capability.yaml, then reload
(/capabilities → reload) — a single comma-joined string is read as one unknown
domain and silently reverts to all 90.
| Domain | Shipped | Adds |
|---|---|---|
core | ✅ | Project / team / org discovery (keep on) |
work | ✅ | Iterations, teams, capacity |
work-items | ✅ | Boards CRUD, WIQL |
repositories | ✅ | Repos and PRs |
wiki | ✅ | Wiki pages |
pipelines | ✅ | Build / release runs, logs, artifacts |
test-plans | — | Test plans and suites (+9 tools) |
search | — | Cross-org search |
advanced-security | — | GHAS alerts via advsec_get_alerts / advsec_get_alert_details (+2) — the only way to read them here, and they need GHAS licensed on the org |
WIQL is the SQL-shaped DSL for Boards; @project and @me are useful built-ins:
SELECT [System.Id], [System.Title], [System.State] FROM workitems
WHERE [System.TeamProject] = @project AND [System.WorkItemType] = 'Bug'
AND [System.State] <> 'Closed' ORDER BY [Microsoft.VSTS.Common.Priority]
https://learn.microsoft.com/azure/devops/boards/queries/wiql-syntax
wit_query_by_wiql (or
wit_my_work_items) returns IDs → wit_get_work_items_batch_by_ids hydrates them.wit_get_work_item, id=…).System.AreaPath uses backslashes — escape them in JSON ("Contoso\\Auth\\Identity").
Don't invent area paths / iterations; list them first.advsec_get_alerts.Pre-engagement reconnaissance workflow using HackerOne MCP tools to enumerate program scope, study prior disclosures, and identify high-value targets. Use when starting a new HackerOne program engagement, building an asset inventory, or planning target selection before active testing.
Pre-submission eligibility check for bug bounty findings. Catches ineligible patterns, AI/scanner false positives, and generates impact justification for borderline lows. Use before writing a report or when assessing whether a finding is worth reporting.
Static code analysis for DOM-based vulnerabilities in client-side JavaScript -- source/sink enumeration via grep and AST tools, data flow tracing, sanitization assessment, and framework-specific sink detection. Use when performing pre-commit reviews, auditing large codebases without dynamic execution, or triaging minified code for XSS issues.
Create custom jxscout analyzers (regex, derived, or script-based) and retrigger analysis. Use when the user wants to find specific code patterns across all project files, add new match kinds, or extend jxscout's static analysis capabilities.
Use when the user wants to connect to Jira, Confluence, or Compass — search issues with JQL, read or create issues and pages, link content across products, or set up Atlassian auth. The connector — not security tradecraft.
Use when the user wants to connect to GitHub, search or read issues / PRs / code, create or update issues and PRs, inspect Actions runs, or review Dependabot / code-scanning alerts. The connector — not security tradecraft.