The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands like grep, test, and wc to automate the analysis of specification and implementation files. These tools are used to detect ambiguity, identify TODOs, and check for specific markers within the codebase.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes untrusted data from user-provided files. Ingestion points: Specification documents (PRDs, SDDs), implementation code, and design files provided for validation across various modes. Boundary markers: The skill lacks explicit delimiters or instructions to treat file content as passive data, which could allow malicious instructions within those files to influence agent behavior during analysis. Capability inventory: The agent has the ability to execute shell commands (grep, test, wc) and read arbitrary files as part of its validation logic. Sanitization: There is no evidence of sanitization or validation of input file paths or contents before they are processed by the agent or used in shell commands, which could lead to command injection if paths are maliciously constructed.

specification-validation