The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of shell utilities such as find, grep, wc, mkdir, and mv to analyze project metrics, search for configuration patterns, and reorganize files. These actions are limited to the local filesystem.
[CREDENTIALS_UNSAFE]: A core feature of the skill is detecting exposed secrets (passwords, API keys, tokens) in CLAUDE.md and other configuration files using regular expression searches. The skill processes these credentials to migrate them to a secure .credentials.json file listed in .gitignore, which is a security best practice for managing secrets in this environment.
[PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it reads and processes user-controlled content from CLAUDE.md and root-level scripts to perform refactoring and organization.
Ingestion points: Reads CLAUDE.md, project root files (*.py, *.sh, etc.), and .claude/ directory contents.
Boundary markers: No explicit delimiters or instructions are used to separate untrusted data from the agent's logic when reading files.
Capability inventory: The skill has the ability to write to the filesystem, move files (mv), and execute shell commands.
Sanitization: No specific sanitization or validation logic is applied to the content of the files before they are processed by the agent.
[DATA_EXPOSURE]: The skill accesses sensitive local paths, specifically the agent's internal project memory directory located at ~/.claude/projects/, to verify the presence and structure of the memory system.

project-optimizer