Skip to content

Support URL/domain exclusions in link validation #68

@blva

Description

@blva

Problem Statement

When running skill-validator check or validate links in CI, some links are structurally valid and intentional but fail with HTTP 404 or connection errors because of the environment — not because the links are broken. Two concrete cases:

  • Private GitHub repos (e.g. https://github.com/<private>/<private>) — return 404 from any runner without org access. The link is correct; CI just can't reach it.
  • Localhost URLs (e.g. http://localhost:8080/api/atlas/v2/openapi) — valid in the context of local dev instructions inside a skill, but always fail in CI.

Currently the only workaround is to skip link validation entirely (validate structure + analyze content + analyze contamination instead of check), which means losing link checks for everything else.

Proposed Solution

A way to exclude specific URLs or domains from link validation, either via:

  • A config file (.skill-validator.yml) with an ignore-links list, e.g.:
    links:
      ignore:
        - "https://github.com/10gen/**"
        - "http://localhost:**"
  • Or a CLI flag: --ignore-link-pattern "github.com/10gen" (repeatable)

Alternatives Considered

  • RFC 6570 template syntax is automatically skipped, but converting real links to templates loses the actual URL which reduces skill quality.
  • Skipping validate links entirely in CI — works as a blunt workaround but removes all external link checking.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions