Build Lean SaaS cube logoBuild Lean SaaS
Back to Skills
lineartriageLinearHermesCodex SkillsAgent Triage

Linear X Bookmark Agent Triage

An installable Hermes/Codex-style skill for reviewing Linear issues created by X Bookmark Capture Sync and turning each capture into a Shape Up Linear task.

Install the skill
Free

The source is public. Install it yourself, or use DevelopJoy when you want help wiring it into Codex, Claude, OpenClaw, Linear, cron, and your repo.

Terminal
mkdir -p ~/.codex/skills/linear/x-bookmark-agent-triage/scripts
curl -fsSL https://buildleansaas.com/api/skills/linear/x-bookmark-agent-triage/files/SKILL.md -o ~/.codex/skills/linear/x-bookmark-agent-triage/SKILL.md
curl -fsSL https://buildleansaas.com/api/skills/linear/x-bookmark-agent-triage/files/scripts/x-bookmark-agent-triage.mjs -o ~/.codex/skills/linear/x-bookmark-agent-triage/scripts/x-bookmark-agent-triage.mjs
chmod +x ~/.codex/skills/linear/x-bookmark-agent-triage/scripts/x-bookmark-agent-triage.mjs
Free skill: source and install command included.
Supported: pair-programming setup help is available through DevelopJoy.
Pair-programming support keeps your agent workspace current
Austin Witherow
3 min read

Installable Source

The website page is the guide. The install command downloads the curated source files from Build Lean SaaS. Maintainers can also copy skills/linear/x-bookmark-agent-triage from the repository root.

This is the second skill in the X bookmark workflow.

The capture skill stays deterministic: fetch bookmarks, dedupe locally, create raw Linear Triage issues.

This skill handles the next step: read those raw Linear issues, prepare them for an agent review pass, and apply the agent's decisions back to Linear with an audit trail.

What it does

  • Finds Linear issues created by x-bookmark-capture-sync.
  • Exports review candidates as JSON for Hermes, Codex, or another agent runner.
  • Generates a review prompt that asks the agent to use Shape Up thinking.
  • Accepts a decisions JSON file from the agent.
  • Adds a Linear comment explaining the decision.
  • Rewrites the issue title and description for converted or question-needed tasks.
  • Tracks processed issue IDs in local JSON state.

Install

Use the install command on this page to copy the public skill files into your local skills directory.

For OpenClaw-style repositories, keep the same folder shape and point the tool at the installed SKILL.md.

Configure

Set these environment variables:

Optional:

Fetch candidates

Fetch raw captures from Linear:

The script looks for issues in your team whose description contains:

The output is intentionally JSON so another agent can score and decide what to do.

Generate the prompt for Hermes, Codex, or another runner:

Agent decision shape

Ask Hermes, Codex, or your runner to read review-prompt.md and write:

Allowed actions:

  • convert: turn the capture into a real task with a better title/description.
  • question: leave clarifying questions as a comment.
  • park: mark it reviewed but intentionally leave it as a parked capture.

Allowed work types:

  • blog_post
  • small_experiment
  • implementation_task
  • research_note

Every decision includes Shape Up fields: problem, appetite, pitch, rabbit holes, no-gos, and next action. The point is not to skip Linear. Everything remains a Linear task; Shape Up decides how the task should be handled.

Apply decisions

Preview first:

Then apply:

V1 does not change status, assignee, labels, or priority. It only comments, optionally rewrites title/description, and records processed issue IDs locally.

Cron or heartbeat

The intended heartbeat has two stages:

That keeps the script deterministic and leaves the reasoning step to an agent that can explain its decisions.