The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and follow instructions embedded within the data it processes from the local project environment.
Ingestion points: The skill executes openspec instructions apply which returns JSON fields named context and operationGuidance. These fields contain project-specific facts and advisory guidance that the agent is instructed to follow.
Boundary markers: The skill explicitly instructs the agent to treat these inputs as separate from the CLI-returned state, ensuring they do not permit bypassing blocked states or replace built-in instructions.
Capability inventory: The skill uses Bash(openspec:*) to execute commands and has the ability to read project files and update task checkboxes in markdown artifacts.
Sanitization: The instructions include a requirement for the agent to report conflicts if the project-provided context or guidance contradicts the core workflow or user choices.
[COMMAND_EXECUTION]: The skill heavily relies on executing the openspec CLI tool with various subcommands (status, instructions, list, store list). Arguments such as --change and --store are dynamically populated based on user input or inferred from the conversation context.

openspec-apply-change