code-context
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
git status --shortand executes repository-defined scripts (e.g., test, build, lint) to establish truth from the project environment. These operations are restricted to existing toolchains defined within the target repository. - [PROMPT_INJECTION]: The skill processes untrusted content from local repository files (e.g.,
README.md,AGENTS.md, and source code), which creates a surface for indirect prompt injection. - Ingestion points: The skill systematically reads various manifests, guidance files, and source entry points throughout the local repository.
- Boundary markers: The agent is instructed to look for repository-defined rules in
AGENTS.mdand mark unchecked areas as 'Not verified', though no explicit prompt-level delimiters are provided for ingested text. - Capability inventory: The agent has capabilities to read local files and execute commands already present in the project's own configuration.
- Sanitization: Repository content is analyzed as evidence without specific sanitization or escaping mechanisms before being processed by the agent.
Audit Metadata