The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/sync-brand-to-tokens.cjs uses child_process.execFileSync to execute a local design token generator located at .claude/skills/design-system/scripts/generate-tokens.cjs. This is an expected workflow to convert brand guidelines into functional CSS tokens.
[COMMAND_EXECUTION]: The regression test file scripts/tests/test_sync_brand_to_tokens.py uses subprocess.run to invoke node and execute the synchronization script. This is a standard testing pattern for verifying script behavior during development.
[INDIRECT_PROMPT_INJECTION]: The skill implements a brand context injection mechanism in scripts/inject-brand-context.cjs that reads docs/brand-guidelines.md and interpolates its content into a system prompt. This constitutes a potential injection surface if the guidelines file is modified to include malicious instructions.
Ingestion points: docs/brand-guidelines.md (read via fs.readFileSync).
Boundary markers: The generated prompt uses headers (e.g., BRAND CONTEXT:) but lacks explicit 'ignore embedded instructions' delimiters for the extracted text.
Capability inventory: The skill has capabilities for local file writes (fs.writeFileSync) and local command execution (execFileSync).
Sanitization: The extraction logic uses specific regex patterns to target hex codes, typography stacks, and table structures, which partially mitigates the risk of full-text instruction obedience.

brand