The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from various project documents which could serve as a vector for indirect prompt injection.
Ingestion points: The skill reads from design/gdd/*.md, design/narrative/, design/levels/*.md, and design/art/art-bible.md (SKILL.md, Phase 1).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat this external content as untrusted data or to ignore embedded commands.
Capability inventory: The skill has the ability to write files (Write, Edit) and spawn sub-agents (Agent) which execute complex instructions based on the ingested data.
Sanitization: No sanitization or validation of the text content extracted from design documents is performed before it is used to generate prompts or specifications.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes dynamic context injection to execute a local shell command when the skill is loaded.
Evidence: The line !`bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config --keys review_mode,automation` triggers a shell command execution to resolve configuration settings.
Context: This execution occurs silently at load time, although in this instance, it targets a local project script for configuration purposes rather than external or sensitive resources.

asset-spec