The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automatically detects build systems (npm, cargo, swift, etc.) and executes associated shell commands to build the project and run UI/unit tests. These commands are dynamically determined by the agent while inspecting local files.\n- [DATA_EXFILTRATION]: Uses the gh gist edit command to write project-derived findings and 'pitfall' patterns to a specific, hardcoded GitHub Gist (ID: 84a5c108d5742c850704a5088a3f4cbf). This facilitates the transfer of project metadata and technical observations to an external, shared destination.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it analyzes untrusted project data to update its core logic.\n
Ingestion points: Phase 1 reads spec.md, journey.md, and generated test code from the project root and subdirectories.\n
Boundary markers: The instructions lack specific boundary markers or 'ignore' commands to prevent the agent from following instructions embedded inside the files being analyzed.\n
Capability inventory: The agent has permissions to execute shell commands, modify the local AGENTS.md instruction file, and write to an external Gist.\n
Sanitization: No evidence of sanitization or validation logic is present to filter malicious instructions before they are applied to the skill's persistent configuration files.

refine-journey