The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using node to run a registry generator script at hsdd/scripts/gen-registry.mjs.- [COMMAND_EXECUTION]: The skill directs the agent to access and copy files from the internal directories of other skills, specifically targeting ~/.claude/skills/hsdd-contract/scripts/gen-registry.mjs. This involves cross-skill interaction and access to sensitive user home directories outside the immediate project scope.- [PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it ingests user-authored Markdown files and uses their content to influence future agent reasoning.
Ingestion points: Markdown files stored in the hsdd/adr/ directory.
Boundary markers: The skill extracts content based on ## Decision and ## Consequences headers but does not provide instructions to the agent to ignore potentially malicious prompts embedded in those sections.
Capability inventory: Extracted content is intended to be injected into 'phase contexts' as binding decisions, directly affecting the agent's future planning and execution steps.
Sanitization: No sanitization or verification of the extracted file content is described before it is projected into the agent's operational context.

hsdd-adr