The Agent Skills Directory

[SAFE]: The skill provides a robust framework for project governance and documentation management without exhibiting malicious patterns.
[COMMAND_EXECUTION]: The skill involves the execution of local scripts scripts/setup_project.py and scripts/validate-project.mjs. These scripts are self-contained, use standard libraries, and are restricted to managing project documentation and configuration files (package.json, CI workflows, and instruction files).
[DATA_EXFILTRATION]: No network activity or data exfiltration vectors were detected. The skill operates entirely within the project root on the local file system.
[PROMPT_INJECTION]: The instructions do not contain any safety bypass or behavior override patterns. They focus on maintaining project state and following established maintenance rules.
[INDIRECT_PROMPT_INJECTION]: The skill features a surface for indirect prompt injection as it processes project data to derive terminology and generate wiki content.
Ingestion points: scripts/setup_project.py reads package.json; the agent is instructed to explore the codebase in SKILL.md Steps 1 and 9.
Boundary markers: The skill employs comment markers for managing its own instruction blocks but relies on workflow separation for ingestion.
Capability inventory: The skill can write to the local file system (docs/, scripts/, CI configs) and modify instruction files.
Sanitization: Risks are effectively mitigated by a mandatory human-in-the-loop workflow. Workflow Step 3 requires explicit agreement from the project owner on derived terms, and Step 9 requires explicit approval of exploration findings before publishing to the wiki.

setup-project