The Agent Skills Directory

[SAFE]: The skill's core functionality is restricted to reading local codebase structures and generating technical documentation.
[DATA_EXPOSURE]: The skill identifies and reads project manifest files (e.g., package.json, pyproject.toml, Cargo.toml) and configuration files to understand project dependencies and architecture. All findings are synthesized into a single markdown report saved locally within the project directory; no instructions for network exfiltration or remote data transmission are present.
[COMMAND_EXECUTION]: The skill relies on standard file-system reconnaissance and reading capabilities. It includes explicit safety rules (Rule 9) that forbid modifying, creating, or deleting files in the target codebase, with the sole exception of writing the final analysis report.
[PROMPT_INJECTION]: The instructions provide a methodical 5-phase execution strategy and do not contain patterns attempting to bypass safety filters or override the agent's core instructions.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface as it ingests untrusted data by reading files from a user-specified directory ({TARGET_DIR}).
Ingestion points: Files throughout the codebase provided as input.
Boundary markers: None explicitly defined to separate project content from the agent's instructions.
Capability inventory: Limited to file system reads and writing a single markdown document. No network access or shell command execution is requested.
Sanitization: None specified for the content read from files. Given the strict instruction to generate a technical report and the lack of sensitive capabilities like network access, the risk is minimal.

codebase-analyzer