The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs users to run local scripts using npx tsx to test hook activation (e.g., .claude/hooks/skill-activation-prompt.ts).
[COMMAND_EXECUTION]: Troubleshooting instructions include the use of chmod +x to make bash wrappers executable within the project's hook directory.
[EXTERNAL_DOWNLOADS]: The guide recommends executing npm install within the .claude/hooks directory to set up the execution environment for TypeScript hooks.
[PROMPT_INJECTION]: The system architecture includes a 'UserPromptSubmit' hook that injects context based on user prompts, which serves as a surface for indirect prompt injection.
Ingestion points: The UserPromptSubmit hook (documented in HOOK_MECHANISMS.md) ingests user prompts to match against patterns in skill-rules.json.
Boundary markers: Output is formatted with graphical separators to distinguish injected context from user input.
Capability inventory: The architecture supports injecting context into the prompt and blocking file-editing tools (PreToolUse hook).
Sanitization: No explicit sanitization or input validation logic for user prompts is described in the implementation guide.

skill-developer