The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill reads ~/.claude/.mcp.json and ~/.claude/settings.json. These files are standard locations for storing Model Context Protocol (MCP) server configurations and agent settings, which frequently contain sensitive information including API keys, bearer tokens, and private environment variables.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute shell commands (wc -w, ls, head) to inspect the local filesystem, count words in files, and list directory contents within the ~/.claude directory.
[DATA_EXFILTRATION]: The instructions contain hardcoded absolute paths for a specific user (C:/Users/gurusharan.gupta/). While intended for local auditing, reading configuration and memory files into the agent's context window makes this sensitive data available to the model, which could be exploited if the model is subsequently instructed to send its context externally.
[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by reading the contents of SKILL.md files from all installed skills and various memory markdown files.
Ingestion points: Reads ~/.claude/skills/*/SKILL.md and ~/.claude/projects/*/memory/*.md using Read and Bash tools.
Boundary markers: None identified; the content of these files is processed directly to generate a report.
Capability inventory: The skill has access to Read, Glob, and Bash tools.
Sanitization: No sanitization or validation of the ingested file content is performed before processing.

context-budget