The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses high-priority override markers such as "CRITICAL" and "ACTIVATION-NOTICE" to command the agent to disregard external files and strictly follow the provided persona definition. These instructions are designed to lock the agent into a specific state of being.
[COMMAND_EXECUTION]: The skill invokes various CLI tools including git, node, supabase-cli, and railway-cli. It utilizes complex shell commands via wsl on Windows to execute Linux binaries like coderabbit.
[DYNAMIC_EXECUTION]: Local scripts including unified-activation-pipeline.js and codebase-mapper.js are executed during the agent's lifecycle to perform activation and project mapping.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data which could contain malicious instructions.
Ingestion points: Reads files from docs/stories/, .aiox/handoffs/, and project codebase content via git logs and diffs.
Boundary markers: Absent; untrusted content is interpolated directly into the agent context without clear delimiters.
Capability inventory: Execution of local scripts via node and shell commands via various CLI tools.
Sanitization: Absent; the skill does not explicitly sanitize or validate the content of the project files it processes.

aiox-architect