The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use a CLI tool named paseo to run and manage background agent processes. This tool is used to execute specific roles (proposers and adjudicators) as part of the architecture review workflow, utilizing configuration-driven parameters for providers and models.
[PROMPT_INJECTION]: The workflow involves an adjudicator agent processing input from two independent proposer agents. This architecture introduces a potential surface for indirect prompt injection. However, the skill mitigates this by defining a COUNCIL_AGENT_BOUNDARY prompt that enforces role constraints and by using structured YAML/Markdown templates for all generated artifacts, which helps maintain separation between instructions and data.
[SAFE]: A shell script (test-paseo-root-council.sh) is included for verifying the skill's structural integrity. The script performs read-only operations using grep to ensure that required architectural triggers and role definitions are present in the skill files and project documentation.

architecture-council