The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the openspec CLI tool via shell commands to list changes, retrieve implementation instructions, and check task status. These operations are scoped to the openspec tool as defined in the allowed-tools frontmatter.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external files (contextFiles) and CLI outputs (context, operationGuidance) that could contain instructions intended to influence the agent's behavior. This presents a vulnerability surface for indirect prompt injection.
Ingestion points: JSON fields returned by openspec instructions apply (specifically context and operationGuidance) and the content of files specified in the contextFiles array (such as proposal, specs, or design documents).
Boundary markers: The skill instructions provide logical boundaries, explicitly telling the agent that external guidance is prompt-level advice that must not replace built-in instructions, evidence of task completion, or the controlling CLI-reported state.
Capability inventory: The skill allows the agent to execute openspec CLI commands, read project files, and modify local files to record task completion.
Sanitization: The instructions direct the agent to report conflicts between external data and the controlling workflow and to avoid copying runtime context verbatim into persistent artifacts.

openspec-apply-change