The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external data from PRDs and issues. Ingestion points: PRDs and issue descriptions provided by users in the working context. Boundary markers: Absent; there are no instructions to the agent to disregard embedded directives within these documents. Capability inventory: The skill can execute shell commands (tests, typechecking), use development tools (/tdd, /review), and perform file system modifications (commits). Sanitization: Absent; the skill does not filter or validate the content of the requirements before processing them.
[PROMPT_INJECTION]: The skill dynamically reads and obeys commands from a project file. Ingestion points: docs/agents/loops.md. Boundary markers: Absent. Capability inventory: The skill executes 'verification commands' defined in the documentation. Sanitization: Absent; the agent is instructed to run commands found in the local file without pre-execution validation.

implement