decomp-code-rules
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill introduces a vulnerability surface by requiring the agent to ingest and act upon instructions from an external project profile.
- Ingestion points: The agent is directed to read the
AGENTS.mdfile from the project directory to retrieve configurations and execution commands. - Boundary markers: There are no delimiters or instructions provided to separate the untrusted project profile from the agent's core instructions.
- Capability inventory: The skill explicitly instructs the agent to run "commands" defined in the project profile, such as rule checkers, per-function checks, and byte gates, which typically involve shell execution.
- Sanitization: No validation or sanitization logic is described for the commands retrieved from the external configuration file, allowing for potential command injection via a malicious project profile.
Audit Metadata