The Agent Skills Directory

[COMMAND_EXECUTION]: The function discover_installed_skills in lib/workflow_state/parity.py executes the command npx skills ls -g --json to identify installed skills and their file paths for parity verification.
[EXTERNAL_DOWNLOADS]: The use of npx to execute the skills utility may result in the download of the package from the official npm registry if it is not cached in the local environment.
[REMOTE_CODE_EXECUTION]: The skill employs importlib.util in lib/workflow_state/inventory.py and scripts/manage_subfeatures.py to dynamically load and execute local Python modules from computed file paths, facilitating integration between planning and execution components.
[PROMPT_INJECTION]: The skill processes untrusted data from project Markdown files and JSON metadata, presenting an indirect prompt injection surface.
Ingestion points: Content is read from Markdown tables in slice-traceability.md and structured fields in .subfeature-meta.json and registry.json.
Boundary markers: Absent. The skill parses these files directly into internal workflow models without specific delimiters to ignore embedded instructions.
Capability inventory: Includes the ability to create directories, write metadata and documentation files, and execute local shell commands.
Sanitization: The skill validates subfeature IDs against a regex pattern and restricts status transitions to a predefined whitelist of valid states.

add-subfeature