The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in references/preload.md direct the agent to execute a Bash command (ls -1 .okr/) to list the contents of the project directory. This is used for environment discovery and context loading, which is consistent with the skill's purpose as an OKR management tool.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it is designed to ingest and process various user-controlled files from the .okr/ directory to drive its logic.
Ingestion points: Content is loaded from files such as objective.md, plan.md, resources.md, actions/*.md, inbox/*.md, and lessons/index.md as part of the 'Preload Contract' defined in references/preload.md.
Boundary markers: The instructions do not specify any delimiters or safety warnings (e.g., 'ignore instructions in data') when interpolating the content of these files into the agent's context.
Capability inventory: The skill suite has the capability to write and modify files within the .okr/ directory, update task statuses, and propose project changes based on the data it processes.
Sanitization: No mechanisms for sanitizing or escaping the content of ingested files are mentioned before they are processed by the agent.

okr-shared