The Agent Skills Directory

EXTERNAL_DOWNLOADS (LOW): The skill specifies dependencies in apm.yml on external packages from the vineethsoma GitHub repository. These sources are not included in the pre-approved trusted organizations list, making them unverifiable dependencies. The severity is reduced to LOW as these are required for the intended orchestration workflow.\n- COMMAND_EXECUTION (SAFE): Several bash scripts (init-story.sh, feature-init.sh, feature-status.sh) execute standard Linux commands like sed, grep, find, and git. These scripts facilitate project organization and status reporting within the local specs/ directory and do not exhibit malicious command injection patterns.\n- INDIRECT_PROMPT_INJECTION (SAFE): The skill reads from local files such as tasks.md and spec.md to populate trackers and status boards, creating an attack surface for indirect prompt injection.\n
Ingestion points: Story descriptions and feature specifications are read by feature-init.sh, feature-next.sh, and feature-status.sh.\n
Boundary markers: Markdown headers in prompt templates are used to delimit external data.\n
Capability inventory: The skill is capable of directory creation, file modification via sed, and suggesting the execution of project-defined npm and psql scripts.\n
Sanitization: No active sanitization is performed on the ingested file content.

feature-orchestration