The Agent Skills Directory

[SAFE]: The skill provides a high-level procedural framework for AI-assisted engineering. It promotes best practices such as defining security boundaries, proposing architectural changes before implementation, and requiring human checkpoints for material decisions.
[DATA_EXPOSURE]: The workflow involves reading and writing to ~/.ai-tools/implementation-notes.md. The skill demonstrates a strong security posture by explicitly instructing the agent to omit secrets, credentials, customer data, and sensitive incident details from these records.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external data (repository evidence and implementation notes), it mitigates risks through a workflow that requires human review and verification of all generated changes and decisions.

accountable-engineering