The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/independent_flow_gate.py uses subprocess.run to execute the codex CLI. This involves executing external binaries based on user-provided arguments (e.g., --codex-bin), which is a sensitive operation that could be abused if the environment is not secured.
[DYNAMIC_EXECUTION]: The instructions in SKILL.md and several scripts reference and execute code located in relative parent directories (for example, node ../seenry/scripts/pick.mjs, python3 ../seenry/scripts/palette.py, and python3 ../seenry-motion/scripts/video_frames.py). This creates a dependency on an external filesystem state and could lead to unauthorized code execution if the surrounding environment is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external mobile applications and user journeys via research tools like search_app_flows and get_app_flow to inform its design tasks.
Ingestion points: Data from external apps enters the agent's context through the Seenry MCP tool outputs described in the Workflow and Research sections.
Boundary markers: The instructions do not define clear delimiters or include warnings for the agent to ignore potentially malicious embedded instructions in the researched app content.
Capability inventory: The skill possesses capabilities including command execution (via the codex CLI), file writing in the workspace, and image generation via external scripts.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content retrieved from external app flows before it is used to drive the agent's logic and design decisions.

seenry-apps