.lycheeignore

# Lychee link-check exclusions. One regex per line (matches anywhere in URL).

# Vendored upstream baselines (plugins/*/skills/*/vendor/) are byte-exact
# snapshots kept solely for drift diffing. Their relative links may point at
# upstream sibling files that are deliberately not vendored — do not check
# links that resolve into a vendor directory.
/plugins/[^/]+/skills/[^/]+/vendor/

# Claude Code expands these runtime variables after loading an installed
# plugin. In source Markdown they are intentionally not filesystem paths.
CLAUDE_(PLUGIN_ROOT|PLUGIN_DATA|PROJECT_DIR|SKILL_DIR)

# Synthetic ai-briefing eval fixtures use a fictional release feed. The URLs
# are test data for ranking/deduplication behavior, not external dependencies.
github\.com/meridian-labs/agent-cli/releases/tag/

# Product UI on claude.ai blocks automated clients: 403 even with a browser
# User-Agent (verified 2026-08-21). Docs on code.claude.com stay checked.
^https?://claude\.ai/

# ACM Digital Library blocks automated clients outright: 403 even with a
# browser User-Agent (verified 2026-08-21). Recurring — new DOI citations
# keep arriving; host-level so the weekly report does not re-file the next one.
^https?://dl\.acm\.org/

# Reddit blocks automated clients: 403 even with a browser User-Agent
# (verified 2026-08-21). Community citations stay in the docs.
^https?://(www\.)?reddit\.com/