linnet-config-customization

name	linnet-config-customization
description	This skill should be used when helping someone use, configure, or customize Linnet, especially for setup, source selection, language/model changes, prompt overrides, sink enablement, or workflow schedule changes.

Linnet Config & Customization

Use this skill when the task is mainly about helping someone set up or customize their own Linnet repository.

Do not use it as the main guide for contributor-facing repo edits. For implementation work inside the repo, prefer skills/linnet-contributor/SKILL.md.

Read order

Start with the user-facing setup path first:

1. README.md or README_zh.md
2. setup-bridge/README.md
3. dev_docs/manual-config.md
4. config/sources.yaml
5. config/extensions/*.yaml for the relevant source
6. sinks/README.md plus the relevant sink README.md if delivery changes are involved
7. .github/workflows/daily.yml, weekly.yml, or monthly.yml if schedule changes are requested

Read references/config-map.md when you want a compact config map.

Explain the config in this order

1. display_order controls the rendered section order
2. each source block uses enabled
3. language controls the output language
4. llm.* controls provider label, API base URL, secret name, model IDs, and prompt overrides
5. sinks.* controls optional delivery channels
6. workflow files control automation timing and all cron expressions are UTC

Rules to preserve

• Keep secrets in environment variables or GitHub Actions secrets only
• Do not move credentials into committed YAML
• Keep optional features clearly optional
• Prefer editing existing config keys over inventing new top-level config
• Keep setup docs and wizard wording aligned with the actual generated config
• Prefer Use this template plus the upstream setup wizard over Fork for first-time users
• Treat the GitHub App + Linnet Bridge flow as the default happy path
• Treat manual config as a fallback path, not the main onboarding story

Common tasks

LLM changes

• provider presets live in the setup wizard
• llm.base_url and llm.api_key_env already support OpenAI-compatible endpoints
• the chosen secret name must match the environment variable that actually exists
• Step 5 deploy instructions should reflect the resolved provider config, not assume OpenRouter

Setup and deploy changes

• the default path is: Use this template -> install Linnet Bridge GitHub App -> authorize browser -> deploy
• do not reintroduce PAT requirements into the default wizard path
• if a user cannot install the GitHub App or is blocked by org policy, move them to the manual guide explicitly

Source changes

• enable or disable sources in config/sources.yaml
• adjust detail config in config/extensions/<name>.yaml
• keep the default user story simple unless the user explicitly wants advanced customization

Sink changes

• enable sinks under sinks:
• keep webhook URLs and tokens out of the repo
• check the sink-specific README for required env var names

Schedule changes

• edit the GitHub Actions workflow cron lines directly
• explain UTC clearly when discussing run times

Validation

Use the smallest useful check for the scope:

• review YAML structure and indentation
• run targeted tests or diagnostics when behaviour changed
• run cd astro && npm run check if setup copy or generated wizard content changed
• mention anything you could not verify

Done means

• the user can see which file or wizard step they need to touch
• secret handling remains safe
• docs match the real current setup flow
• advanced options stay discoverable without complicating the default path