The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands like ripgrep, git log, and perl to perform repository-wide audits and inventory matching lines of prose against specific regex patterns.
[COMMAND_EXECUTION]: The skill provides instructions to install a shell script located at scripts/commit-msg into the .git/hooks/commit-msg directory. This establishes a persistence mechanism where the script is executed automatically by the operating system during every future git commit operation.
[PROMPT_INJECTION]: The audit mode functionality (Step 2
Catalog) is vulnerable to indirect prompt injection. It reads raw snippets from untrusted files in the working directory and commit history into the agent's context. These snippets are presented in a table without sanitization or explicit boundary markers, potentially allowing malicious content within project files to influence the agent's behavior. Evidence chain: Ingestion points: working directory files and commit logs via rg and git log; Boundary markers: none; Capability inventory: ripgrep, git, install, perl, sed; Sanitization: none.

negafix