Linear Native Node

Version: 1.0.22 / local Linear utility.

Use scripts/linear.mjs to talk directly to Linear's GraphQL API at https://api.linear.app/graphql.

Native Node 18+. Zero npm dependencies. Windows, macOS, and Linux friendly. No bash, Python, OAuth gateway, browser login, or third-party proxy.

Security behavior

• Reads LINEAR_API_KEY and optional LINEAR_DEFAULT_TEAM from the process environment only.
• Does not read credential files or ~/.openclaw/.env.
• Sends GraphQL requests only to https://api.linear.app/graphql.
• Does not write local files, log secrets, or print the API key.
• Linear personal API keys inherit the permissions of the user/account that created them. Treat output as workspace-sensitive; use the least-privileged key model Linear offers if scoped keys become available.
• Read-only commands are the default usage pattern.
• Write commands are explicit commands (create, comment, status, priority, project-create) and are fail-closed by default: the script exits unless --execute is present after explicit user approval for that specific Linear mutation.
• --execute is only a script-level flag gate. Mutation appropriateness is enforced by the operator/agent approval policy: do not run a write unless the current user instruction or approval names the Linear action, target, and intended mutation.
• Static-analysis note: env-var + network-send warnings are expected for this skill because it uses a named API key to call Linear's fixed GraphQL endpoint. Treat any broad env enumeration, dynamic endpoint, or key-printing behavior as a blocker.
• Read command output may include workspace names, user names, emails, issue titles, issue descriptions, and URLs from the connected Linear workspace. Redact that content before sharing logs or examples externally.

Why this skill

This is the local/native Node version of common Linear helper workflows:

• Keeps API access between this machine and Linear only.
• Uses LINEAR_API_KEY from the process environment only; never print the key.
• Avoids bash-oriented helpers that break on Windows.
• Avoids third-party OAuth/gateway tools or intermediary services.
• Keeps a tight auditable scope: viewer/org/team lookup, issue list/read/create/update/comment, practical personal summaries, and branch names.

Setup

Create a Linear API key at:

https://linear.app/settings/api

PowerShell session-only setup:

$env:LINEAR_API_KEY = "<your-linear-key>"
$env:LINEAR_DEFAULT_TEAM = "TEAM"

macOS/Linux shell session-only setup:

export LINEAR_API_KEY="<your-linear-key>"
export LINEAR_DEFAULT_TEAM="TEAM"

LINEAR_DEFAULT_TEAM is optional. Examples below use TEAM and TEAM-123 as placeholders; replace them with your Linear team key and issue identifiers.

Run

From PowerShell, bash, zsh, or any shell with Node on PATH:

node "<skill-dir>/scripts/linear.mjs" help
node "<skill-dir>/scripts/linear.mjs" viewer
node "<skill-dir>/scripts/linear.mjs" issues --team TEAM --mine --limit 10

From OpenClaw, use the bundled script directly:

Run: node <skill-dir>/scripts/linear.mjs my-issues --team TEAM
Run only after explicit approval: node <skill-dir>/scripts/linear.mjs --execute create TEAM "Fix login error" "Users see an error after submitting credentials." --priority high

Add --json before the command when structured output is useful. Global flags (--json, --execute) are intentionally prefix-only; put them before the command so positional text such as a comment body of --execute or --json is not consumed as a global flag.

Commands

Read-only:

help
viewer
organization
teams
projects [--team TEAM] [--limit N]
states <TEAM_KEY>
issues [--team TEAM] [--mine] [--state STATE] [--limit N]
my-issues [--team TEAM] [--state STATE] [--limit N]
my-todos [--team TEAM] [--limit N]
urgent [--team TEAM] [--mine] [--limit N]
standup [--team TEAM] [--limit N]
issue <IDENTIFIER>
branch <IDENTIFIER>

For list commands, --team TEAM narrows results to one team. If --team is omitted and LINEAR_DEFAULT_TEAM is set, that default is used where supported. If neither is set, list commands that support broad lookup query across the accessible workspace scope instead of failing. states <TEAM_KEY> always requires a team key.

Writes to Linear require explicit approval and --execute:

--execute project-create [TEAM_KEY] "Name" ["Description"]
--execute create [TEAM_KEY] "Title" ["Description"] [--priority urgent|high|medium|low|none]
--execute comment <IDENTIFIER> "Comment"
--execute status <IDENTIFIER> <state-name>
--execute priority <IDENTIFIER> <urgent|high|medium|low|none>

For project-create and create, TEAM_KEY may be omitted only when LINEAR_DEFAULT_TEAM is set; with two or more positional args, the first arg is treated as the explicit team key. Without prefix --execute, write commands fail before contacting Linear.

For agent use, execute write commands only when the user's current instruction explicitly names the Linear action, target, and intended mutation (for example, "create a Linear issue for..." or "move TEAM-123 to Done") or the user explicitly approves the exact command.

Examples

List teams, projects, and states:

node "<skill-dir>/scripts/linear.mjs" teams
node "<skill-dir>/scripts/linear.mjs" projects --team TEAM --limit 20
node "<skill-dir>/scripts/linear.mjs" states TEAM

Create a project:

node "<skill-dir>/scripts/linear.mjs" --execute project-create TEAM "Release Planning" "Track launch scope and milestones."
# If LINEAR_DEFAULT_TEAM is set:
node "<skill-dir>/scripts/linear.mjs" --execute project-create "Release Planning"

List issues:

node "<skill-dir>/scripts/linear.mjs" issues --team TEAM --limit 20
node "<skill-dir>/scripts/linear.mjs" issues --team TEAM --mine --state "In Progress"
node "<skill-dir>/scripts/linear.mjs" my-issues --team TEAM
node "<skill-dir>/scripts/linear.mjs" my-todos --team TEAM
node "<skill-dir>/scripts/linear.mjs" urgent --team TEAM

Read and update an issue:

node "<skill-dir>/scripts/linear.mjs" issue TEAM-123
node "<skill-dir>/scripts/linear.mjs" --execute comment TEAM-123 "Added reproduction notes."
node "<skill-dir>/scripts/linear.mjs" --execute status TEAM-123 "In Progress"
node "<skill-dir>/scripts/linear.mjs" --execute priority TEAM-123 urgent

Create an issue:

node "<skill-dir>/scripts/linear.mjs" --execute create TEAM "Fix notification delivery" "Notifications are delayed for some users." --priority high
# If LINEAR_DEFAULT_TEAM is set:
node "<skill-dir>/scripts/linear.mjs" --execute create "Fix notification delivery" --priority high

Generate a GitHub-friendly branch name from the Linear issue title:

node "<skill-dir>/scripts/linear.mjs" branch TEAM-123

Get machine-readable output:

node "<skill-dir>/scripts/linear.mjs" --json my-issues --team TEAM

Notes

• Prefer issue identifiers like TEAM-123; the script resolves them before updates that need Linear IDs.
• Priority values map to Linear's native values: urgent, high, medium, low, none.
• Missing keys, 401/403 responses, proxy/egress denial hints, GraphQL errors, missing teams/states, and invalid priorities produce clear stderr errors with non-zero exit codes.
• When omitting TEAM_KEY via LINEAR_DEFAULT_TEAM, pass only the project name or issue title. Descriptions require an explicit TEAM_KEY because two positional arguments are parsed as TEAM_KEY plus name/title.
• List commands default to --limit 25; standup defaults to --limit 20. Limits must be integers from 1-100. The script returns the first page ordered by Linear updatedAt; it does not auto-paginate.
• Linear HTTP and GraphQL errors surface as clear stderr errors with non-zero exit codes, but response bodies/details are omitted because they may contain workspace-sensitive data; the script does not retry or back off. For 401/403, verify both Linear key permissions and whether an egress proxy/sandbox is blocking api.linear.app.
• standup summarizes assigned issues in started and completed Linear workflow states within the returned first page; it is intentionally lightweight, not a time tracker or date-windowed report.
• Help is available with help, -h, or --help. There is no --version flag; use the version: frontmatter and changelog.

Sample output

Sanitized representative output for eval/review checks:

Example success output

$ node scripts/linear.mjs viewer
Example User <user@example.com>

$ node scripts/linear.mjs my-issues --team TEAM --limit 2
TEAM-123 [In Progress/high] @Example User - Fix notification delivery
TEAM-124 [Todo/medium] @Example User - Draft release checklist

Example error output

$ node scripts/linear.mjs create TEAM "Fix login error" "Users see an error after submit."
error: write command "create" is blocked by default. Re-run with --execute only after explicit approval.

$ node scripts/linear.mjs viewer
error: LINEAR_API_KEY is not set.
Create a Linear API key at: https://linear.app/settings/api
Then add it for this shell session:
  PowerShell: $env:LINEAR_API_KEY = "<your-linear-key>"
  bash/zsh:    export LINEAR_API_KEY="<your-linear-key>"
Optional default team:
  PowerShell: $env:LINEAR_DEFAULT_TEAM = "TEAM"
  bash/zsh:    export LINEAR_DEFAULT_TEAM="TEAM"

Changelog

• 1.0.22: Frontmatter wording hygiene: clarify that the skill has no third-party intermediary/proxy service while still calling Linear's own GraphQL API directly. No runtime Linear behavior change and no ClawHub/public publish authority.
• 1.0.21: Test/source-review hygiene: construct the fake Linear API key sentinel in tests.mjs instead of embedding a credential-looking literal, so source-attached review packaging can scan the test harness without weakening secret gates. No runtime Linear behavior change and no ClawHub/public publish authority.
• 1.0.20: Public-clear packaging polish: exclude dry-runs/ review evidence from ClawHub packages, move current dry-run evidence out of the skill source tree, and extend package leak tests to scan expected packaged files for Windows path variants and review/evidence inventory markers. No categories, topics, topic tags, tags, keywords, or ClawHub catalog metadata added to source.
• 1.0.19: Source polish: remove remaining direct/direct wording, make global flags prefix-only so positional text equal to --execute/--json is not consumed as a global flag, omit workspace-sensitive Linear response bodies/details from non-auth HTTP/GraphQL errors, and clarify write/JSON examples. No categories, topics, topic tags, tags, keywords, or ClawHub catalog metadata added to source.
• 1.0.18: Metadata wording hygiene: remove "Direct" from the description; no runtime behavior change.
• 1.0.17: Superseded local listing-topic experiment removed before final retest; no runtime behavior change.
• 1.0.16: Public-wording hygiene patch: remove ambiguous current publishability/public-readiness language from the body/changelog so source text cannot be mistaken for standing ClawHub/public approval; no runtime behavior change.
• 1.0.15: Version refresh after prior packaging/public-readiness review; no runtime behavior change and no standing publish approval implied.
• 1.0.14: Reject extra positional arguments for single/two-argument commands, including write commands, to prevent truncated operator intent.
• 1.0.13: Add cross-platform missing-key hints, clarify 401/403 proxy-vs-auth diagnostics, guard mutually exclusive state filters, and expand offline gate tests.
• 1.0.12: Clarify that --execute is a script-level flag gate while mutation appropriateness remains an operator/agent approval-policy requirement.
• 1.0.11: Tighten public docs for Linear key permissions, team/default-team behavior, write approval wording, pagination/rate-limit behavior, help support, sample-output scanability, and consistent --limit validation.
• 1.0.10: Add sanitized sample outputs for eval review, rename the write-gate helper for clearer fail-closed call sites, and keep local test harness excluded from ClawHub packages.
• 1.0.9: Security/audit polish: document the write-command approval gate and expected static-analysis warning for env-var + fixed Linear endpoint usage. No functional API behavior change.
• 1.0.8: Add fail-closed --execute gate for Linear write commands and align setup examples.
• 1.0.7: Package metadata cleanup: restore display name and tags after package-only cleanup.
• 1.0.6: Package-only cleanup: publish runtime files plus packaging metadata (SKILL.md, .clawhubignore, and scripts/linear.mjs) and keep the local test harness out of the public archive.
• 1.0.5: Fix GraphQL variable construction in issue list queries (issues, my-issues, my-todos, urgent, standup) so only actually used variables are declared and passed.

Changelog tracked here from 1.0.5 onward.