The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's summary.mjs script uses execFileSync to run local git commands. This is used solely to capture short commit hashes for stamping the generated documentation. The execution uses hardcoded arguments and targets a local project subdirectory, minimizing the risk of command injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes local project files, such as markdown specifications and progress reports, to generate its reading aids. This represents an ingestion surface for untrusted data if those files are modified by external parties. However, the risk is mitigated by the skill's specific task-oriented design and its use of sanitization for the resulting output.
Ingestion points: Files located within the hsdd/management/, hsdd/spec/, hsdd/contract/, and hsdd/adr/ directories are read by various extraction scripts.
Boundary markers: The skill does not employ explicit delimiters for the agent when reading these files, but it operates through a structured CLI-driven extraction and validation workflow.
Capability inventory: The skill possesses capabilities for reading and writing files within the project structure and executing local git commands.
Sanitization: Robust sanitization is present in the views-core.mjs and stamp.mjs files, which handle HTML escaping and safe JSON serialization to prevent injection into the generated offline documents.
[SAFE]: The skill vendors the @dagrejs/dagre library (v3.1.1) for diagram layout. This inclusion is accompanied by clear documentation in README.md, legal notices, a source URL pointing to the NPM registry, and a SHA256 hash for integrity verification, adhering to standard secure vendoring practices for third-party dependencies.

hsdd-summary