The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes instructions and content from the openspec CLI and project files to generate artifacts. This data is treated as authoritative without explicit boundary markers or sanitization, creating a surface for indirect prompt injection.
Ingestion points: Data from openspec status, openspec instructions, and openspec context commands, as well as existing dependency files.
Boundary markers: The skill does not use specific delimiters or 'ignore embedded instructions' warnings to isolate content from these sources.
Capability inventory: The skill can execute openspec commands and write files to the filesystem.
Sanitization: No sanitization or validation is performed on the data received from the CLI before it is used to generate output or influence the agent's actions.
[DYNAMIC_EXECUTION]: The skill supports dynamic command or skill execution when the openspec instructions output delegates artifact creation to a specific command. This pattern allows external CLI data to direct the agent to execute specific actions.
[COMMAND_EXECUTION]: The skill executes various openspec CLI commands to manage the change proposal workflow, including new change, status, instructions, and context.

openspec-propose