The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local auditing script (scripts/audit.sh) to verify project settings. The script utilizes standard system tools like grep, find, and jq for file inspection.
[SAFE]: The skill checks for project registration in local developer tool configurations, such as obsidian.json and afk.json. These operations are performed locally and do not involve credential theft or data exfiltration.
[SAFE]: Evaluation of the indirect prompt injection surface:
Ingestion points: Project documentation and configuration files (e.g., AGENTS.md, package.json) and local developer environment configs.
Boundary markers: Audit results are structured into a deterministic markdown table.
Capability inventory: Limited to read-only operations and pattern matching using shell utilities.
Sanitization: Uses deterministic tools (grep, jq) for verification, ensuring that project content is treated as data rather than instructions.

repo-standard