The Agent Skills Directory

[COMMAND_EXECUTION]: The provided Python script scripts/codebase_analyzer.py performs local file system operations to scan directory structures and file metadata. It uses only the Python standard library and does not execute arbitrary shell commands or external binaries.
[DATA_EXFILTRATION]: The script is designed to capture repository signals such as file counts, detected languages, and top-level structures. It intentionally ignores sensitive directories like .git, node_modules, and virtual environments (venv, .venv). It does not read file contents, only metadata like file names and sizes.
[CREDENTIALS_UNSAFE]: The documentation templates in references/output-format-templates.md follow security best practices by recommending the use of environment variables (process.env.NOTION_TOKEN, $CONFLUENCE_TOKEN) rather than hardcoding API keys or secrets.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted data (local codebases), it only extracts structural metadata and uses standard formatting (JSON or plain text) for its output. There is no automated path for this data to be executed as instructions by the agent.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns were found. The skill does not perform external downloads or install unverified packages.

codebase-onboarding