The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected during the analysis. The skill's architecture relies on local execution of audited scripts and enforces a clear separation of authority between auditing and fixing components.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from repository context files which could potentially contain malicious instructions. However, the skill implements multiple robust safeguards: it uses specific JSON-only response templates with clear delimiters, enforces a verbatim quotation contract for extracted claims, and applies a mandatory citation gate that verifies all evidence against the actual repository tree. These structural controls effectively mitigate the risk of a malicious file influencing the agent's behavior.
[COMMAND_EXECUTION]: The skill executes local Git commands and Node.js scripts to perform its analysis. These operations are strictly scoped to the target repository and the skill's internal runtime. Command arguments are passed as arrays to child processes, which prevents shell injection, and the tool performs extensive path normalization to ensure operations do not escape the intended directory.

context-audit