The Agent Skills Directory

[SAFE]: The skill is explicitly identified as a 'v8-conformant fixture' intended for testing purposes. Its content is limited to YAML frontmatter configuration and markdown documentation explaining how to use it within a development environment.
[COMMAND_EXECUTION]: A command example (node scripts/skill-lint.js examples/fixture-skills) is provided in the documentation for verification purposes. This is an instructional example for the user to run a local linter and does not represent an automated or hidden execution pattern.
[INDIRECT_PROMPT_INJECTION]: While the skill defines relationship predicates (related, suppresses, verify_with, depends_on) that influence agent routing, these are static configuration fields following a strict schema. The ingestion point is the platform's own skill-loading mechanism, and no untrusted external data is processed.
[METADATA_POISONING]: The audit-state.json file contains metadata regarding the skill's versioning and verification status. The fields are standard for a testing suite and do not contain deceptive instructions or executable payloads.

with-relations