The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates by ingesting and analyzing user-provided plans, design documents, and existing codebase artifacts to identify unresolved decisions. This pattern creates a surface for indirect prompt injection, where malicious instructions could be embedded within the processed data to influence the agent's logic.
Ingestion points: Processes external user input and workspace files as directed in SKILL.md and references/codebase-first.md.
Boundary markers: The skill does not include instructions for using specific delimiters or 'ignore' instructions when reading untrusted content.
Capability inventory: The skill utilizes file-reading and local search tools (grep, git log) to analyze the environment, as documented in references/codebase-first.md.
Sanitization: No explicit sanitization or validation of the ingested external content is defined within the skill's workflow.

plan-interview