The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands using the openspec CLI tool, including openspec new change, openspec status, and openspec instructions. These commands are used to manage project state and retrieve metadata.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and follows natural language instructions and rules provided dynamically by the output of the openspec instructions command.
Ingestion points: The skill reads JSON data containing context, rules, and instruction fields from the output of openspec instructions --json as defined in SKILL.md.
Boundary markers: Absent; there are no specific delimiters or warnings used to isolate the instructions provided by the tool from the agent's core system instructions.
Capability inventory: The agent has the ability to execute shell commands (openspec) and perform file system operations (creating and writing markdown files) across all steps in SKILL.md.
Sanitization: Absent; the agent is explicitly instructed to apply the retrieved context and rules as constraints for its output without validation.

openspec-propose