The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external sources such as brain-dumps, PRDs, RFCs, and design documents to generate system specifications. This architectural pattern constitutes an indirect prompt injection surface, as the agent processes untrusted external data that is then interpolated into generated spec files. However, the risk is mitigated by the highly structured nature of the output (markdown specifications rather than executable code).
Ingestion points: Root brain-dumps, external product requirement documents (PRD), request for comments (RFC), and design tickets (SKILL.md, Section 1: Normalize the input).
Boundary markers: The skill uses specific markdown templates and section headers to organize data, but does not explicitly instruct the agent to ignore instructions embedded within the source documents.
Capability inventory: The skill primarily performs file creation and manipulation within the project's dedicated directory (hsdd/spec/) and references other internal skills (hsdd-contract, hsdd-adr).
Sanitization: No specific sanitization or filtering of text from source documents is described.
[COMMAND_EXECUTION]: The project conventions file (templates/conventions.md) includes instructions for the user or agent to execute local maintenance commands, such as 'node hsdd/scripts/gen-registry.mjs' and 'openspec init'. These are standard project initialization and registry generation tools relevant to the HSDD ecosystem.
[EXTERNAL_DOWNLOADS]: The skill contains references to external documentation URLs and recommends third-party companion tools hosted on GitHub (e.g., github.com/obra/superpowers). These are used for documentation and enhancing the developer workflow and do not involve silent or untrusted code execution.

hsdd-spec