The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an 'Interview' phase that ingests untrusted user input which is subsequently used to generate a PRD, Implementation Plan, and Task List. This creates a surface for indirect prompt injection where malicious input could influence the content of generated files or the behavior of the downstream /implement-feature command.
Ingestion points: User input collected during the 'Requirement Gathering (Interview)' step (Step 1) in SKILL.md.
Boundary markers: Absent; the instructions do not define delimiters or specific warnings to ignore instructions embedded within the user-provided requirements during PRD generation.
Capability inventory: The skill performs file writes to the local workspace (docs/specs/, implementation_plan.md, task.md) and executes the /implement-feature command.
Sanitization: No sanitization, escaping, or validation of the gathered user requirements is described before they are interpolated into the workflow.
[NO_CODE]: The skill consists exclusively of markdown instructions and metadata. It does not include any standalone scripts, binaries, or configuration files that execute logic outside of the agent's prompt context.

plan-feature