The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a custom command rota block decisions in Step 7. This command is used to update managed blocks within project configuration files like AGENTS.md or CLAUDE.md, ensuring that documented decisions are integrated into the agent's operational context. This represents expected functionality for the skill's purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources and conversations to update agent instructions, which presents a surface for potential indirect prompt injection.
Ingestion points: Data is read from the active conversation, .rota/KNOWLEDGE.md, and files within .rota/spikes/ (referenced in SKILL.md Step 2 and source-prefill-modes.md).
Boundary markers: Step 5 of the skill (Confirmation Gate) requires the user to manually review the full draft and explicitly select "Write it" before any persistent changes are made.
Capability inventory: The skill has the ability to write to .rota/DECISIONS.md, modify AGENTS.md or CLAUDE.md, and execute the rota shell command.
Sanitization: The skill relies on the user confirmation gate for validation rather than automated sanitization of the ingested content.

rota-decide