The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes the exarchos_orchestrate tool to retrieve 'runbooks'—sequences of execution steps—at runtime (e.g., the task-completion and agent-teams-saga runbooks). The orchestrator is then directed to execute these returned steps, representing dynamic script generation and execution.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data in the form of implementation plans and review reports, which are interpolated into subagent prompts.
Ingestion points: External implementation plans and review results are read via exarchos_workflow get and processed in SKILL.md and references/fix-mode.md.
Boundary markers: While the skill uses markdown headers to separate task descriptions, it does not implement strict sanitization or escaping of the interpolated plan content.
Capability inventory: The skill possesses the capability to spawn subagents (Task primitive), execute shell commands (git, npm, test runners), and write to the filesystem.
Sanitization: The instructions do not specify any sanitization, filtering, or validation of external content before it is placed into subagent or fixer prompts.
[COMMAND_EXECUTION]: The skill orchestrates environment setup and task execution through numerous shell commands, including git worktree management, npm install for dependencies, and project-specific testing tools. Subagents are specifically instructed to perform cd operations into worktrees based on paths supplied by the orchestrator.

delegate