The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform system-level operations for repository analysis, including line counting using find and wc, and contributor analysis using git log and git shortlog commands.
[DATA_EXFILTRATION]: The skill is explicitly instructed to search for and read sensitive files such as .env, .env.local, and other credential or secret files within the target codebase. While this is a core part of its technical due diligence function to assess security risk, it involves accessing sensitive file paths that could be abused if the agent is redirected to a host's sensitive directories.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its primary function is to ingest and analyze large amounts of untrusted external data (a codebase) without using security boundary markers.
Ingestion points: The agent reads all source files, documentation (README.md, ARCHITECTURE.md), and configuration files from the target repository using the Read, Glob, and Grep tools.
Boundary markers: None. The instructions do not direct the agent to wrap external content in delimiters or to ignore instructions that might be embedded in code comments or documentation.
Capability inventory: The skill has access to powerful tools including Bash (shell execution), Read (file system access), and the ability to write a final report to the local file system.
Sanitization: None. The skill processes and samples raw text from the codebase directly into the model's context, creating a surface where malicious code comments could influence agent behavior.

tech-due-diligence