The Agent Skills Directory

[PROMPT_INJECTION]: No instructions were found that attempt to override agent constraints or bypass safety filters. All prompts are dedicated to the technical writing process and agent coordination.\n- [DATA_EXFILTRATION]: Access to the codebase is used solely for documentation purposes. While the skill extracts information from configuration and environment files (e.g., .env.example), there are no network operations targeting external domains for exfiltration.\n- [REMOTE_CODE_EXECUTION]: Use of the Bash tool is limited to local Git commands (git log, git tag, git diff) to determine project status and history. There is no evidence of downloading or executing scripts from remote servers.\n- [COMMAND_EXECUTION]: Shell command execution is properly scoped to repository analysis and follows the stated purpose of the skill without exposing arbitrary command injection surfaces.\n- [SAFE]: The architecture uses a multi-agent approach with strict role definitions, input/output contracts, and a quality review stage (critic-agent), which aligns with best practices for secure and reliable agent design.

docs-writing