The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the openspec CLI and project files, creating a surface for indirect prompt injection. A malicious project could use the context or operationGuidance fields to attempt to override agent instructions.
Ingestion points: The agent ingests data from openspec status --json, openspec instructions apply --json, and every file listed in the contextFiles array in SKILL.md.
Boundary markers: The skill provides explicit instructions to the agent to treat context and operationGuidance as advisory only, ensuring they do not replace built-in instructions or bypass blocked states.
Capability inventory: The skill utilizes the Bash(openspec:*) tool to execute CLI commands and performs file modifications during the implementation phase in SKILL.md.
Sanitization: There is no automated sanitization of the input; the skill relies on the agent's reasoning to identify and report conflicts between project guidance and the controlling workflow.
[COMMAND_EXECUTION]: The skill uses the openspec CLI to perform various actions like listing stores, checking status, and getting instructions. These calls are scoped to the openspec tool as defined in the allowed-tools frontmatter.

openspec-apply-change