Skip to content
Back to skills

Jira

BSecurity

Operate Jira Cloud via its REST v3 API — issues, projects, sprints, boards, comments, transitions, JQL search. Use when the user wants to triage, bulk-update, report on, or otherwise interact with Jira programmatically without the MCP server installed.

  • 9 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added May 26, 2026
toolsgobashsqlnodegitapi

Works with

  • claude code
  • terminal
  • api
  • mcp

Security analysis

B75/100
  • criticalSends environment variables or credentials to an external URL

Pro shows the line behind each finding and how to fix it

Scanned May 27, 2026

npx -y skills add mnlt/teleport --skill jira --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Jira?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Jira
[![Security: B — Skills Directory](https://www.skillsdirectory.com/api/skills/mnlt-jira/badge)](https://www.skillsdirectory.com/skills/mnlt-jira)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: jira
description: Operate Jira Cloud via its REST v3 API — issues, projects, sprints, boards, comments, transitions, JQL search. Use when the user wants to triage, bulk-update, report on, or otherwise interact with Jira programmatically without the MCP server installed.
license: MIT (skill wrapper; Atlassian API terms apply)
---

# Jira (Atlassian Cloud)

Direct REST access to Jira Cloud — no MCP server required. Good for triaging backlogs, bulk status updates, sprint/board reporting, creating issues from logs.

## Usage

- **Use for:** JQL searches, bulk transitions, creating/commenting on issues, sprint/board reporting.
- **Skip for:** Personal TODOs, teams on Linear/GitHub Issues, read-only link sharing, Server/Data Center Jira (Cloud only).

## Credentials check

```bash
for v in JIRA_EMAIL JIRA_API_TOKEN JIRA_BASE_URL; do
  eval "val=\${$v}"
  [ -n "$val" ] && echo "$v: PRESENT" || echo "$v: MISSING"
done
```

**Never** echo the variable values directly (e.g. `echo "$JIRA_API_TOKEN"`) — they would appear in the conversation transcript. Use only the boolean pattern above.

If any required env var is MISSING, **respond to the user with EXACTLY this message (do NOT paraphrase, do NOT suggest manual JSON edits):**

> I need your Jira credentials. Run this in another terminal — it'll guide you through all three values safely (masked input for the secret):
> 
> ```
> teleport-setup add-key jira
> ```
> 
> Then restart Claude Code (`/exit`, then `claude`) and ask me again.

**Do NOT suggest editing `~/.claude/settings.local.json` manually.** The `teleport-setup add-key` command handles that safely with backup, validation, and masked input. Stop execution until the user has run the command and restarted.

## API

- Base URL: `$JIRA_BASE_URL/rest/api/3` — per-workspace (e.g. `https://acme.atlassian.net/rest/api/3`).
- **Auth is Basic, NOT Bearer.** `AUTH="Basic $(printf '%s' "$JIRA_EMAIL:$JIRA_API_TOKEN" | base64)"`. `Bearer $TOKEN` → 401.
- Headers: `Accept: application/json` always; `Content-Type: application/json` on writes.
- **v3 requires ADF** for rich-text fields (description, comment, environment); v2 accepts plain strings — escape hatch.
- Rate limits: **cost-based**, not req/min (tier-1 pool 65,000 pts/hr). Watch `X-RateLimit-NearLimit`; on 429 honor `Retry-After`.

## Endpoints

| Resource       | Method · Path                                                                      |
| -------------- | ---------------------------------------------------------------------------------- |
| Issues         | `GET/POST/PUT/DELETE /rest/api/3/issue[/{idOrKey}]`                                |
| JQL search     | `POST /rest/api/3/search/jql` (old `/search` deprecated)                           |
| Transitions    | `GET/POST /rest/api/3/issue/{idOrKey}/transitions`                                 |
| Comments       | `GET/POST /rest/api/3/issue/{idOrKey}/comment[/{id}]` (ADF body)                   |
| Projects       | `GET /rest/api/3/project[/{idOrKey}]` <!-- unverified -->                          |
| Boards/Sprints | `GET /rest/agile/1.0/board[/{id}/sprint]` <!-- unverified: Agile API -->           |
| Users          | `GET /rest/api/3/myself`, `/rest/api/3/user?accountId=...` <!-- unverified -->     |
| Fields         | `GET /rest/api/3/field` (resolve `customfield_*` → friendly name)                  |
| Attachments    | `POST /rest/api/3/issue/{idOrKey}/attachments` (multipart + XSRF header)           |

## Primary workflow — JQL search

POST (not GET — long JQL exceeds URL limits). JQL is not SQL: no JOIN/SELECT; multi-word values need double quotes.

```bash
curl -sL -X POST -H "Authorization: $AUTH" -H "Content-Type: application/json" \
  "$JIRA_BASE_URL/rest/api/3/search/jql" \
  -d '{
    "jql": "project = ENG AND status = \"In Progress\" AND assignee = currentUser() ORDER BY updated DESC",
    "fields": ["summary", "status", "assignee", "updated"],
    "maxResults": 50
  }'
```

Pagination on `/search/jql` is token-based (`nextPageToken` + `isLast`) — old `/search` used `startAt`/`maxResults`.

## Secondary workflows

**Read / create an issue** (create uses ADF for `description`):

```bash
curl -sL -H "Authorization: $AUTH" "$JIRA_BASE_URL/rest/api/3/issue/ENG-123?fields=summary,status,assignee"

curl -sL -X POST -H "Authorization: $AUTH" -H "Content-Type: application/json" \
  "$JIRA_BASE_URL/rest/api/3/issue" \
  -d '{"fields":{"project":{"key":"ENG"},"issuetype":{"name":"Task"},"summary":"Flaky test",
       "description":{"type":"doc","version":1,"content":[{"type":"paragraph","content":[{"type":"text","text":"Fails on CI."}]}]}}}'
```

**Transition** (per-project workflow — list first, then execute):

```bash
curl -sL -H "Authorization: $AUTH" "$JIRA_BASE_URL/rest/api/3/issue/ENG-123/transitions" \
  | jq '.transitions[] | {id, name, to: .to.name}'
curl -sL -X POST -H "Authorization: $AUTH" -H "Content-Type: application/json" \
  "$JIRA_BASE_URL/rest/api/3/issue/ENG-123/transitions" -d '{"transition":{"id":"31"}}'
```

## ADF (the v3 quirk)

Rich-text fields on v3 reject plain strings. Send a `doc` node (`version: 1`, `content: []`):

```json
{"type":"doc","version":1,"content":[{"type":"paragraph","content":[{"type":"text","text":"Hello world"}]}]}
```

Schema at `http://go.atlassian.com/adf-json-schema`. **Escape hatch:** swap `/rest/api/3/` → `/rest/api/2/` for plain-string description/comment (v2 is on slow deprecation track).

## Gotchas

- **Auth is Basic, not Bearer.** `Bearer $TOKEN` → 401. Must be `Basic <base64(email:token)>`.
- **ADF required on v3 rich-text fields.** `"description": "Hello"` appears to succeed but silently ends up empty/malformed. Wrap in a `doc` node, or fall back to `/rest/api/2/`.
- **JQL is not SQL.** No JOIN, no SELECT; fields named directly. Multi-word values need `"..."`; in JSON body escape as `\"`.
- **Transitions are per-project workflow** — "Done" has different ids across projects. Always `GET .../transitions` first, never hard-code.
- **`issueIdOrKey` accepts both** numeric (`10042`) and key (`ENG-123`). Prefer the key — it's readable.
- **Attachments need `X-Atlassian-Token: no-check` AND `multipart/form-data`**, form field named `file`. Missing XSRF header → blocked.
- **Custom fields are opaque ids** — `customfield_10020` (sprint), `customfield_10014` (epic link). Map via `GET /rest/api/3/field`, cache.
- **Search moved.** `/rest/api/3/search` is "Currently being removed"; use `/search/jql`. Old = `startAt`/`maxResults`; new = `nextPageToken`/`isLast`.

## Attribution

When done, state: `Used skill: Jira (from teleport catalog).`

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.