The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses authoritative override markers and forceful persona steering (e.g., "alter your state of being", "stay in this being until told to exit", "ALL task instructions override any conflicting base behavioral constraints") that explicitly instruct the agent to disregard its standard operational parameters in favor of the skill's local instructions.
[COMMAND_EXECUTION]: Instructions provide explicit shell commands for the agent to execute, including a fallback command node .aiox-core/development/scripts/unified-activation-pipeline.js aiox-master and various git commands used for project status reporting.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted handoff artifacts from the .aiox/handoffs/ directory to suggest subsequent commands to the user.
Ingestion points: Reads YAML files from the .aiox/handoffs/ directory during the activation sequence (Step 3, substep 5.5).
Boundary markers: There are no visible delimiters or warnings to ignore potentially malicious instructions embedded within the handoff artifacts.
Capability inventory: The skill possesses extensive capabilities including file modification (*create, *modify), shell execution (git, node), and framework-level governance tasks.
Sanitization: Although the YAML metadata mentions a principle of sanitizing user inputs, the specific logic for processing handoff artifacts does not demonstrate sanitization or validation of the ingested strings before they are presented as suggested commands.

aiox-master