The Agent Skills Directory

[COMMAND_EXECUTION]: The skill lifecycle involves executing shell commands to initialize and manage session data, specifically using mkdir and redirection (cat >) to create directory structures and context files within the .council/ directory as outlined in references/SESSIONS.md.
[INDIRECT_PROMPT_INJECTION]: The skill's core functionality relies on subagents researching topics via external tools and then feeding that information back to a central coordinator. This creates a surface for indirect prompt injection.
Ingestion points: Specialist subagents ingest data from untrusted external sources via WebSearch and Context7 tools as described in SKILL.md and references/SPECIALISTS.md.
Boundary markers: The skill uses markdown templates (found in references/TEMPLATES.md) to structure councillor analysis, but it lacks strict data/instruction delimiters to prevent malicious external content from influencing the coordination logic.
Capability inventory: The skill environment supports shell command execution for file management and subagent orchestration through task-based tools.
Sanitization: The instructions do not specify any sanitization or validation of the data retrieved from external sources before it is synthesized by the "Chair" agent.

the-council