The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands for project builds, installations, and testing (e.g., npm, cargo, make) as part of its task-completion logic. These operations are expected for its role as an automation agent and are categorized as safe within that context.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and processes user stories from files like prd.json and progress.txt. Evidence chain: 1. Ingestion points: prd.json and progress.txt files. 2. Boundary markers: Absent. 3. Capability inventory: Shell command execution, file writing, and agent delegation. 4. Sanitization: Absent. This surface is inherent to the skill's purpose and is mitigated by mandatory reviewer verification.

ralph