The Agent Skills Directory

[DATA_EXFILTRATION]: The skill exposes the host's internal file structure and a specific username ('nicol') through hardcoded absolute paths in the execution_context and process sections.
[PROMPT_INJECTION]: The skill is designed to ingest and act upon data from project files (STATE.md, CONTEXT.md, and .continue-here), creating a surface for indirect prompt injection.
Ingestion points: Reads project status and context from local markdown and state files.
Boundary markers: None. The skill does not provide delimiters or instructions to ignore potential commands within these files.
Capability inventory: The skill can spawn new agents and request user input based on the contents of the ingested files.
Sanitization: None. Content from project files is used directly to drive the agent's workflow logic.
[PROMPT_INJECTION]: The skill includes a risk: safe metadata tag, which is a deceptive instruction intended to bypass safety assessments.
[COMMAND_EXECUTION]: The skill delegates its primary processing logic to an external file (resume-project.md) located at a fixed local path. This execution pattern relies on content that is not bundled with the skill and cannot be verified for safety or integrity.

gsd-resume-work