skills.rest
JaviMontano
JaviMontanoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

deep-link-runbook

Designs and audits claude-cli://open deep links with prompt prefill and workspace targeting.

Updated Jun 28, 2026
Search Tags:#onboarding#claude code#runbook#claude-cli#url encoding#deep link#prompt prefill
One-click install
npx skills add https://github.com/JaviMontano/claude-plugins --skill deep-link-runbook-javimontano
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-link-runbook
Source: https://github.com/JaviMontano/claude-plugins/tree/main/plugins/claude-native-toolkit/skills/deep-link-runbook
Command: npx skills add https://github.com/JaviMontano/claude-plugins --skill deep-link-runbook-javimontano

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Creating Claude Code deep links by hand is error-prone: URLs need correct encoding, the cwd versus repo precedence is easy to get wrong, and teams often embed secrets or wrongly assume the prompt auto-executes. This Skill produces safe, well-formed claude-cli://open links and audits existing ones for contract, privacy, and platform issues. ## Core Features & Use Cases - Deep Link Construction: Builds encoded claude-cli://open URLs using only the supported q, cwd, and repo parameters, enforcing the 5,000-character prompt limit and cwd precedence rules. - Safety and Platform Audits: Reviews links for embedded secrets, PII, or sensitive logs, and covers handler registration, browser prompts, GitHub Markdown stripping, and fallback instructions. - Surface Decision Guidance: Determines when a deep link is the wrong tool and recommends headless claude -p, scripts, channels, hooks, scheduled tasks, or Remote Control instead. - Use Case: A team lead needs an onboarding link that opens Claude Code in a new developer's cloned repo with a read-only prompt asking for a five-step checklist; the Skill outputs the encoded link, a GitHub-safe code-block fallback, and a privacy review. ## Quick Start Ask the assistant to create a claude-cli://open onboarding link for your local repository path with a read-only prompt and a GitHub-safe fallback.

Frequently Asked Questions about deep-link-runbook

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a Claude Code deep link that opens with a pre-filled prompt?▼

Build a claude-cli://open URL with the q parameter containing your URL-encoded prompt, optionally adding cwd for a local directory or repo for a GitHub owner/name slug. The prompt appears pre-filled but is not sent until the operator presses Enter.

What is the difference between cwd and repo in Claude Code deep links?▼

The cwd parameter targets an absolute local directory path, while repo takes a GitHub owner/name slug resolved to a prior local clone. When both are present, cwd takes precedence even if the path does not exist, so include both only intentionally.

Does a Claude Code deep link run the prompt automatically?▼

No, deep links only pre-fill the prompt box; the operator must review the text and press Enter. They do not approve tools, edit files, or bypass permissions, so use headless claude -p or scripts for non-interactive execution.

Why does my claude-cli:// link not work in a GitHub README?▼

GitHub-rendered Markdown strips unknown protocol links, including custom schemes like claude-cli://. Publish the link inside a code block with a plain-text fallback prompt, or use raw HTML in controlled internal docs instead.

What is the maximum prompt length for a Claude Code deep link?▼

The q parameter has a documented 5,000-character maximum, and Slack display reliability suggests staying near 1,000 characters. For longer instructions, prefill a concise prompt that points to a local runbook file or issue.

When should I use headless claude -p instead of a deep link?▼

Use headless claude -p when CI or scripts need non-interactive execution, machine-readable JSON output, or exit codes. Deep links are for human-reviewed launches like onboarding and runbook kickoff, not automation or event ingestion.

Related Skills

Explore More Software Engineering Skills→
rllm-orgrllm-org
213

hive-setup

Automate Hive agent installation, registration, task cloning, and environment preparation.

Official
Advanced
cosmicstack-labscosmicstack-labs
364

technical-writing

Structure SOPs, user manuals, and release notes with version tracking.

Official
Intermediate
gavinmcfallgavinmcfall
3

config-packager

Audit, sanitize, and package the ~/.claude configuration directory into a distributable repository.

Community
Advanced
tomada1114tomada1114
4

claudecode-headless-automating

Run Claude Code headlessly from shell scripts and CI/CD pipelines.

Community
Intermediate