Review plans, code diffs, and documents through a browser-based annotation CLI.
npx skills add https://github.com/Yassimba/loom --skill plannotator-yassimba
Please help me install this Agent Skill. Skill: plannotator Source: https://github.com/Yassimba/loom/tree/main/skills/plannotator Command: npx skills add https://github.com/Yassimba/loom --skill plannotator-yassimba
plannotator review, with feedback returned as plaintext or JSON. - Document and URL Annotation: Annotate markdown, config files, HTML pages, web URLs, running local apps, or folders with plannotator annotate. - Approval Gates and Guided Reviews: Enforce machine-checkable approval gates with exit codes, browse archived plan decisions, and export or share encrypted Guided Review walkthroughs. - Use Case: After an agent produces a plan, the plan-review hook opens it in the browser; the human annotates concerns, and the structured feedback returns to the agent to address in the same conversation. ## Quick Start Use the plannotator skill to review my current uncommitted code changes and return the reviewer's feedback.High-intent search queries and answers about installing and using this skill.
Run plannotator review followed by the PR URL, such as plannotator review https://github.com/owner/repo/pull/123. It requires an authenticated gh or glab CLI and builds a local checkout of the PR head by default for full file access.
Run plannotator annotate followed by the file path or https URL. Markdown, text, config, and HTML files open directly, while URLs are fetched and converted via Jina Reader unless you pass --no-jina.
Yes, set PLANNOTATOR_REMOTE=1 to force remote mode, which uses fixed port 19432 with a wide bind. You can also publish a session over your tailnet with the --tailscale flag for HTTPS access.
The annotate command deliberately refuses .env files because they commonly hold secrets and annotate history copies file contents. Source-code files belong to plannotator review, which handles code diffs instead.
Add the --json flag to receive one JSON record with decision and feedback fields. For a strict approval gate, combine --gate --json with --require-approval so the exit code reports the human outcome.
ysk1031Generate and revise implementation plans through iterative human annotations.
AutoByteusExecutes one dependency-ready micro-task and records implementation results with evidence.
kaistenbergFetch, verify, and resolve GitHub PR review comments against actual code.
ductrung-nguyenFetch GitHub pull request context and post code reviews with inline comments.