The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands such as git rev-parse, basename, and git commit to manage project metadata and repository state. It interpolates user-provided or filesystem-derived feature slugs into these commands, which could represent a command injection surface if the environment does not sanitize arguments.
[DATA_EXFILTRATION]: The skill is designed to perform a git push to a remote 'origin' repository to sync the vault. This functionality involves transmitting local file content over the network to an external server. Additionally, the use of git add -A stages all changes in the vault, which could inadvertently include sensitive files not related to the specific project.
[PROMPT_INJECTION]: The skill processes untrusted input from prd.md and tech-spec.md files, creating a surface for indirect prompt injection.
Ingestion points: Reads PRD and technical specification files from the local vault.
Boundary markers: Absent; the instructions do not include specific delimiters or warnings to ignore embedded instructions within these documents.
Capability inventory: The skill possesses filesystem read/write capabilities and shell command execution (git).
Sanitization: Absent; the content is parsed directly for task extraction without explicit filtering of potential instructions.

ai-create-tasks