The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The audit script (scripts/audit.mjs) is designed to read environment files and configuration files that may contain sensitive information. Specifically, it reads project-level .env files and global agent settings like ~/.claude/settings.json and ~/.claude.json.- [DATA_EXFILTRATION]: The skill accesses sensitive configuration data from the user's home directory across multiple platforms (Claude, Codex, Cursor, OpenCode). This exposes user-wide agent metadata, active MCP server names, and installed skills to the current project's agent context.- [COMMAND_EXECUTION]: The skill executes a bundled Node.js script (scripts/audit.mjs) to perform its audit. The script utilizes node:fs for deep filesystem access and node:os for gathering system-level environment information.- [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting and parsing the content of external instruction files and rules. Ingestion points: Reads CLAUDE.md, AGENTS.md, and .cursor/rules/*.mdc files during the audit process. Boundary markers: The script does not use specific delimiters or instructions to prevent the agent from following directives contained within the files it audits. Capability inventory: The script performs file reads and calculates token metrics; the results are used by the agent to propose configuration changes. Sanitization: There is no sanitization of the content of the files being audited before they are processed by the script and presented to the agent.

token-saver-opencode