The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes openspec CLI commands using a variable derived from user descriptions. While it instructs the agent to convert input to kebab-case, there is a risk of command injection if the user-provided description contains shell metacharacters that are not properly sanitized during the conversion.
[DYNAMIC_EXECUTION]: The skill writes generated content to an outputPath provided dynamically by the openspec instructions command. This allows the external tool output to dictate where files are written on the local filesystem, potentially leading to arbitrary file writes.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from the openspec CLI output and local dependency files to guide the agent's generation process. This creates a surface where malicious project files could influence the agent's behavior. 1. Ingestion points: openspec instructions JSON output and local dependency artifact files. 2. Boundary markers: The instructions explicitly tell the agent to treat context and rules as constraints and not to include them in the final output. 3. Capability inventory: Execution of shell commands (openspec) and writing files to user-specified paths. 4. Sanitization: The skill does not define explicit technical sanitization for content read from the CLI or files.

openspec-propose