The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill employs the !commandsyntax at the top of `SKILL.md` to execute a shell command:!bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config --keys ... ``. This command runs automatically when the skill is accessed, replacing the placeholder with the script output.
[COMMAND_EXECUTION]: The skill frontmatter and body reference an external script yaml-helper.sh via a path traversal sequence (../../hooks/). This script resides outside the skill's own directory. The allowed-tools section explicitly permits the execution of this specific script with arguments.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from project files to drive its design decomposition workflow.
Ingestion points: Reads design/gdd/game-concept.md, design/game-brief.md, and other project documents in Phase 1.
Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions within the read files.
Capability inventory: The skill uses Write and Edit for file modification, Agent for spawning new sessions, and a scoped Bash tool.
Sanitization: No content sanitization or validation of the ingested markdown data is performed before processing.

map-systems