The Agent Skills Directory

[DATA_EXPOSURE]: The skill is designed to audit sensitive configuration files including ~/.claude/settings.json and project-level memory files (CLAUDE.md). These locations may contain environment variables, path configurations, or tool permissions.
[COMMAND_EXECUTION]: The skill utilizes high-privilege tools such as Bash, Write, and Task to perform audits and coordinate sub-agents like audit-orchestrator and permissions-auditor.
[EXTERNAL_DOWNLOADS]: The metadata installation command fetches the skill definition from a public GitHub repository via curl (raw.githubusercontent.com/vbonk/claude-code-optimizer/main/SKILL.md). This is documented as a reference to a well-known service.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection (Category 8) because it ingests configuration data that may be controlled by third parties in shared repositories.
Ingestion points: Reads CLAUDE.md, .claude/settings.json, and other configuration files in SKILL.md.
Boundary markers: No specific delimiters or instructions are provided to distinguish between audited data and the agent's instructions.
Capability inventory: The skill employs Bash, Write, and Task tools as documented in SKILL.md.
Sanitization: There is no explicit logic for validating or sanitizing the content of the configuration files before processing.

claude-code-optimizer