The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting data from external web searches and incorporating it into the agent's context across multiple research steps (Step 02 through Step 06).
Ingestion points: Web search results for game industry analytics, market trends, and regulatory requirements entering the context in step-02-domain-analysis.md, step-04-regulatory-focus.md, and others.
Boundary markers: Absent; the skill does not use delimiters or explicit instructions to ignore potentially malicious commands embedded in the retrieved web data.
Capability inventory: The agent has the capability to write research content to the file system and execute local Python scripts via subprocess calls.
Sanitization: Absent; the workflow does not include mechanisms to validate, filter, or escape the content retrieved from the web before processing it.
[COMMAND_EXECUTION]: The skill executes a local Python script (resolve_customization.py) to resolve configuration and customization parameters from the project environment.
Evidence: Found in SKILL.md (Step 1) and step-06-research-synthesis.md (On Complete).
Context: The script is a project-local tool used by the vendor to manage workflow settings across multiple skills.
[COMMAND_EXECUTION]: The skill dynamically executes instructions retrieved at runtime from configuration keys such as activation_steps_prepend, activation_steps_append, and on_complete.
Evidence: Described in the activation sequence of SKILL.md and the completion logic of step-06-research-synthesis.md.
Context: This behavior allows for extensibility by executing instructions defined in the user's or team's local .toml configuration files.

gds-domain-research