The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is granted run_shell_command and write_file permissions. It establishes an 'ExecutionProxy Interface' (using the rtk prefix) to run terminal actions such as testing and building, which involves executing shell commands at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external competitor pricing and market data. This creates a surface for indirect prompt injection where untrusted external content could potentially influence the agent's actions, particularly concerning its filesystem and shell access.
Ingestion points: Phase 1 of the Revenue Workflow involves using WebSearch to identify competitor pricing models (SKILL.md).
Boundary markers: None explicitly defined for interpolating external content into the agent context.
Capability inventory: The skill has access to run_shell_command, write_file, replace, and read_file (SKILL.md).
Sanitization: The skill includes a 'Security & Multi-Agent Hygiene' protocol that mandates treating external data as hostile and redacting secrets or PII before sharing context with subagents (SKILL.md).

revenue-architect