The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted configuration files from a project directory. A malicious actor could embed prompt injection instructions within these configuration files (e.g., inside JSON strings, YAML comments, or .env files) to influence the agent's behavior during the validation process.
Ingestion points: The ConfigurationAnalyzer and RuntimeConfigValidator classes read various configuration formats including .json, .yaml, .env, and .js using Path.read_text() and standard parsers in references/detailed-guide.md.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the content of these files as untrusted data or to ignore embedded instructions.
Capability inventory: The skill includes capabilities to write files (implied documentation generation), perform runtime monitoring (chokidar), and execute complex Python/TypeScript logic via subprocesses or internal agent tools.
Sanitization: No sanitization or filtering of the configuration content is evident before it is processed by the agent.
[CREDENTIALS_UNSAFE]: The skill includes logic to scan for secrets (secret_patterns in references/detailed-guide.md). While this is a security feature, the analysis process itself involves reading sensitive files like ~/.env, which could lead to accidental exposure of credentials in the agent's context or logs if not handled carefully.
[COMMAND_EXECUTION]: The Python code in references/detailed-guide.md uses the re module and Path objects to search through project directories. While the provided code is for analysis, agents implementing these instructions may use subprocesses to run the validation scripts, which could be exploited if file paths or configuration values are not properly sanitized.

deployment-validation-config-validate