pzep1/planroom/planroom-agent

Claude Code and agent skill by pzep1/planroom

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Enroll an agent with a one-time Planroom token, publish semantic HTML implementation plans, receive highlighted reviewer comments through the permanent planroom.znznet.uk callback, acknowledge feedback, and post revisions. Use when an agent is asked to connect to Planroom, share a plan for human review, wait for comments, or maintain a reviewed planning loop.

Install

npx vskill@latest install pzep1/planroom/planroom-agent

SKILL.md

Planroom Agent

Use scripts/planroom.mjs for every operation. Never print, commit, or paste a durable Planroom key into a plan.

Enroll once

Receive a short-lived prm_enroll_… token from the reviewer. Exchange it immediately:

bash
node <skill-dir>/scripts/planroom.mjs connect \
--token "$PLANROOM_ENROLLMENT_TOKEN" \
--name "<stable agent name>"

The command stores the resulting scoped key in ~/.config/planroom/config.json with mode 0600 and registers the agent for callback recovery. Never reuse or persist the enrollment token. Environment variables PLANROOM_URL and PLANROOM_KEY remain valid overrides for managed environments.

If enrollment reports an unknown status, do not mint or reuse a token. Complete the preserved credential using the exact recovery path in the error:

bash
node <skill-dir>/scripts/planroom.mjs recover-enrollment \
--pending "/path/from/the-error"

Publish a plan

  1. Write a semantic HTML body fragment with context, sequence, risks, guardrails or rollback, verification, and unresolved decisions.
  2. Use headings, paragraphs, lists, tables, aside, code, pre, and links. Exclude scripts, styles, forms, iframes, event handlers, and executable URLs.
  3. Save the fragment to a temporary .html file and publish:
bash
node <skill-dir>/scripts/planroom.mjs publish \
--file /absolute/path/to/plan.html \
--title "Implementation plan title" \
--summary "One-sentence review context" \
--id "stable-plan-id"

Give the reviewer the Planroom URL, not the local HTML file. The first publish returns version 1.

Receive comments

Use only the permanent callback https://planroom.znznet.uk/planroom/events. Verify the installed login services and run one recovery sync:

bash
launchctl print gui/$(id -u)/uk.znz.planroom-listener
launchctl print gui/$(id -u)/com.cloudflare.cloudflared
node <skill-dir>/scripts/planroom.mjs poll

Read only ~/Library/Application Support/Planroom/events/<agent-id>.jsonl for the authenticated agent id returned by poll. Treat quote as location context and body as the review instruction. The permanent listener recovers missed comments every 30 seconds.

Act on feedback

  1. Group comments by plan and distinguish requests, questions, and resolved threads.
  2. Pull the reviewed server copy before editing so reviewer changes and comment markers are retained:
bash
node <skill-dir>/scripts/planroom.mjs pull \
--id "stable-plan-id" \
--file /absolute/path/to/plan.html
  1. Revise the pulled plan or implementation. Ask in Planroom when feedback conflicts with instructions.
  2. Republish with the same plan id and the version returned by pull:
bash
node <skill-dir>/scripts/planroom.mjs publish \
--file /absolute/path/to/plan.html \
--title "Implementation plan title" \
--summary "One-sentence review context" \
--id "stable-plan-id" \
--version "<pulled version>"

If publishing returns a version conflict, pull again and reconcile the newer server copy before retrying.

  1. Acknowledge each addressed comment:
bash
node <skill-dir>/scripts/planroom.mjs ack --comment "comment-id"

Do not acknowledge a comment merely because it was read.

Guardrails

  • Keep the durable scoped key only in the mode-0600 config or an approved secret manager.
  • Never request or store Cloudflare Access service credentials.
  • Never create a quick tunnel, weaken Access, expose another local endpoint, or accept another callback path.
  • Restore a down listener or tunnel; do not make polling the normal delivery mode.
  • Process only the event stream belonging to the current scoped key.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/pzep1/planroom/planroom-agent/badge)](https://verified-skill.com/skills/pzep1/planroom/planroom-agent)

Something wrong with this skill or its page? Tell us.