The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly executes shell commands retrieved from the verifyCommand field within task metadata. While this is the intended functionality for task verification, it represents a direct execution surface for arbitrary strings stored in data.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests and processes untrusted data from external tasks.
Ingestion points: The TaskGet <task-id> operation in SKILL.md reads task descriptions and metadata into the agent's context.
Boundary markers: The skill relies on parsing json:metadata fences to extract parameters.
Capability inventory: The skill can execute shell commands (verifyCommand), dispatch subagents (subagentBrief), and modify task states (TaskUpdate) across the system.
Sanitization: No sanitization or validation of the verifyCommand or acceptanceCriteria content is described; the agent is instructed to run the provided command and capture exact output.

checking-gates