The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides structured prompt templates (e.g., in references/prompt-patterns.md) designed to strictly control agent behavior according to the 'Spec-Driven Development' methodology. These prompts utilize explicit 'Do NOT' constraints, boundary markers, and 'Always do / Never do' rules to prevent the agent from unilaterally making technical decisions or straying from the defined task scope. These instructions are functional and do not attempt to override safety filters or exfiltrate information.
[DATA_EXFILTRATION]: The skill includes security-focused instructions in its templates (e.g., constitution.md, spec.md) that explicitly forbid the logging or hardcoding of sensitive information like tokens, passwords, and PII. It encourages the use of environment variables and cryptographic hashing for sensitive data storage.
[COMMAND_EXECUTION]: A benign shell helper function (extract-task) is documented in references/ai-agent-patterns.md as an example for extracting specific sections from markdown files using standard system utilities like awk. No malicious command execution or subprocess spawning patterns were detected.
[NO_CODE]: This is a documentation-centric skill providing templates, prompts, and workflow instructions. It does not include any executable scripts, binaries, or automated runtime components. All workflows are executed via the AI agent's reasoning or human-initiated shell commands as documented in the reference files.

spec-driven-development