The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local utility script, scripts/adr.sh, to scaffold ADR files. Analysis of the script shows it is implemented with robust safety practices, including input sanitization via a slugify function and the use of temporary files (mktemp) for atomic writes. It does not perform any network operations or access sensitive system paths.
[DATA_EXFILTRATION]: No network-capable commands (such as curl or wget) or patterns that read sensitive local files (such as SSH keys or environment variables) were identified in the skill instructions or scripts.
[PROMPT_INJECTION]: The SKILL.md file contains detailed instructional guidelines that define a tech-lead persona for the agent. These instructions are focused on software design methodologies and do not contain any patterns intended to bypass safety filters, override system constraints, or extract system prompts.
[REMOTE_CODE_EXECUTION]: The skill is self-contained and does not download or execute remote scripts. It relies on standard internal logic and the included local bash script.
[SAFE]: The skill follows security best practices for its intended purpose, such as recommending secret management (workload identity) and network isolation (private subnets) in its cloud architecture guidelines.

software-architecture