Problem
Coding agents (Copilot CLI agent, Claude, etc.) have no structured way to discover and use gh-devlake commands. Users must manually explain the CLI's capabilities in every agent session. The Playwright CLI solved this with an installable skill definition ΓÇö a structured document that teaches agents the CLI's command surface, typical workflows, and expected output formats.
Proposed Solution
Ship a skills/gh-devlake/SKILL.md (with optional references/ docs) and an install --skills command that copies the skill into the agent's skill directory.
Architecture (following Playwright CLI pattern)
skills/
gh-devlake/
SKILL.md # Top-level command reference + quick start
references/
deployment.md # deploy local, deploy azure, cleanup
configuration.md # connections, scopes, projects, blueprints
querying.md # query commands (from #61), status
troubleshooting.md # diagnose, common errors
SKILL.md structure
Following microsoft/playwright-cli/skills/playwright-cli/SKILL.md:
---
name: gh-devlake
description: Deploy, configure, and query Apache DevLake from the terminal. Manages GitHub, GitLab, and Copilot data sources with DORA metrics.
---
# DevLake Management with gh-devlake
## Quick start
gh devlake deploy local
gh devlake configure full
gh devlake status --json
## Commands
### Deployment
gh devlake deploy local # Docker Compose on this machine
gh devlake deploy azure # Azure Container Apps
gh devlake cleanup # tear down resources
### Configuration
gh devlake configure connection list --json
gh devlake configure connection add --plugin github --token $TOKEN --org my-org
gh devlake configure scope add --plugin github --connection-id 1 --repos org/repo1
gh devlake configure project add --project-name my-team
### Operations
gh devlake status --json # deployment health + connections
gh devlake query dora --project my-team --json
gh devlake insights "question" # AI-powered Q&A
## Output modes
(document --json flag from #60)
## Specific tasks
(link to reference docs)
Install command
gh devlake install --skills
Implementation:
- Detects agent type from environment (check for
.copilot/, .claude/, .github/copilot-instructions.md)
- Copies
skills/gh-devlake/ into the appropriate skill directory
- For Copilot:
.github/skills/gh-devlake/ or .copilot/skills/gh-devlake/
- For Claude:
.claude/skills/gh-devlake/
- Prints confirmation with path
Reference implementation: microsoft/playwright-cli install --skills ΓÇö installs skill files that agents auto-discover.
Skills-less operation (fallback)
Agents can also bootstrap from --help:
gh devlake --help
gh devlake configure --help
gh devlake query --help
This is the Playwright CLI's "skills-less operation" pattern.
Update script
Following microsoft/playwright-cli/scripts/update.js, create a script that regenerates the SKILL.md from the current CLI's --help output and registered commands. This keeps the skill in sync with the implementation.
Acceptance Criteria
Target Version
v0.4.3 ΓÇö Agent-friendly CLI foundation within the active v0.4.x line.
Dependencies
References
Problem
Coding agents (Copilot CLI agent, Claude, etc.) have no structured way to discover and use
gh-devlakecommands. Users must manually explain the CLI's capabilities in every agent session. The Playwright CLI solved this with an installable skill definition ΓÇö a structured document that teaches agents the CLI's command surface, typical workflows, and expected output formats.Proposed Solution
Ship a
skills/gh-devlake/SKILL.md(with optionalreferences/docs) and aninstall --skillscommand that copies the skill into the agent's skill directory.Architecture (following Playwright CLI pattern)
SKILL.md structure
Following
microsoft/playwright-cli/skills/playwright-cli/SKILL.md:Install command
Implementation:
.copilot/,.claude/,.github/copilot-instructions.md)skills/gh-devlake/into the appropriate skill directory.github/skills/gh-devlake/or.copilot/skills/gh-devlake/.claude/skills/gh-devlake/Reference implementation:
microsoft/playwright-cliinstall --skillsΓÇö installs skill files that agents auto-discover.Skills-less operation (fallback)
Agents can also bootstrap from
--help:This is the Playwright CLI's "skills-less operation" pattern.
Update script
Following
microsoft/playwright-cli/scripts/update.js, create a script that regenerates the SKILL.md from the current CLI's--helpoutput and registered commands. This keeps the skill in sync with the implementation.Acceptance Criteria
skills/gh-devlake/SKILL.mdexists with command reference, quick start, and examplesskills/gh-devlake/references/has at least deployment.md, configuration.mdgh devlake install --skillscopies the skill into the agent's directory--helpoutput is sufficient for skills-less operationgo build ./...andgo test ./...passTarget Version
v0.4.3 ΓÇö Agent-friendly CLI foundation within the active v0.4.x line.
Dependencies
--jsonoutput flag to read commands #60 ΓÇö--jsonoutput flag (skill docs reference JSON output mode)References
microsoft/playwright-cli/skills/playwright-cli/SKILL.mdΓÇö the skill file format, command reference structure, and examples sectionREADME.md#installing-skillsΓÇö how agents discover and consume the skillREADME.md#skills-less-operationΓÇö fallback when skills aren't installedscripts/update.jsΓÇö keeps skill content in sync with CLI changesskills/playwright-cli/references/session-management.mdΓÇö pattern for reference docs alongside the main SKILL.mdgh-devlakeAGENTS.md ΓÇö existing architecture documentation to draw from for skill content.github/skills/ΓÇö existing skill definitions in the repo (cli-architecture, plugin-registry, etc.)