The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a project-local script via node hsdd/scripts/gen-registry.mjs. This is a standard part of the project's internal registry maintenance and does not involve external downloads or untrusted sources.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing content from external Markdown files to apply governance updates.
Ingestion points: The skill reads hsdd/spec/*.md and hsdd/conventions.md to find ## Governance updates sections, and it greps the entire project tree for Open Question (OQ) citations.
Boundary markers: The skill relies on specific Markdown headers for structure; however, there are no explicit instruction-level delimiters to prevent the agent from obeying malformed or adversarial instructions within those sections.
Capability inventory: The skill allows the agent to read and write Markdown files, update contract frontmatter, and execute a local Node.js script.
Sanitization: Significant risk mitigation is provided by mandatory human arbitration for all collisions, breaking changes, and contract requests before updates are applied.

hsdd-reconcile